MORTISCOPE: YOLO26 Model Training and Evaluation

Author: Mortiscope

This Jupyter Notebook serves as the core development and training environment for the object detection component of the MORTISCOPE system. The primary objective is to train a YOLO26 model to accurately detect and classify the different life stages of Chrysomya megacephala. The successful training and evaluation of this model are fundamental to the application’s ability to process images and provide accurate entomological data for PMI estimation.

Workflow Summary

This notebook follows a pipeline designed not just for training, but also for deployment and preliminary analysis of forensic imagery:

1. Environment & Storage Initialization: Mounts Google Drive for persistent artifact storage and installs specific dependencies, including sahi for advanced inference.
2. Dynamic Dataset Aggregation: Downloads multiple versioned datasets (varying in resolution and environmental conditions) from Roboflow and programmatically builds a unified data.yaml configuration file to prevent path errors.
3. Training Pipeline: Configures a YOLO26-Large model with custom augmentation strategies. It features a smart resume system that automatically detects interruptions and resumes training from the last saved checkpoint in Google Drive, alongside a custom callback to archive weights for every epoch.
4. Comprehensive Evaluation: Generates training loss plots, a normalized confusion matrix for biological stage differentiation, and detailed inference speed benchmarks.
5. Advanced Inference Setup: Integrates SAHI (Sliced Aided Hyper Inference) to handle high-resolution field imagery to allow the detection of small insect instars that standard resizing would miss.
6. Deployment & Demonstration: Exports the final model to ONNX format for edge deployment and creates an interactive inference engine. This engine detects scene types (Macro vs. Field) to automatically switch between standard and sliced inference modes, applies outlier filtering, and generates publication-ready visualizations.

Tech Stack

Deep Learning and Inference Engines

• Ultralytics YOLO26: The backbone object detection architecture. This notebook specifically uses the Large (l) variant to prioritize high accuracy for subtle morphological features of insect life stages.
• SAHI (Sliced Aided Hyper Inference): An advanced library used to perform inference on high-resolution images by slicing them into smaller overlapping windows. This is critical for detecting small instars in wide-angle “field” shots.
• PyTorch: The underlying tensor computation framework powering YOLO26.
• ONNX (Open Neural Network Exchange): Used to export the trained PyTorch model into a standardized format to optimize it for deployment on various hardware platforms without Python dependencies.

Data Management and Processing

• Roboflow: Used for versioning, hosting, preprocessing, augmenting and downloading the multi-part Chrysomya megacephala dataset.
• Google Drive: Serves as the persistent storage layer to ensure trained weights (best.pt, last.pt) and historical logs are preserved across Google Colab sessions.
• Pandas: Used to parse and manipulate the training logs (results.csv) for custom performance plotting.
• PyYAML: Automates the creation of the dataset configuration file, dynamically linking downloaded image paths to the model.

Visualization and Analysis

• OpenCV (cv2): Handles image I/O, color space conversion, and drawing bounding boxes for the inference demonstration.
• Seaborn & Matplotlib: Used to generate high-quality statistical plots, including the smoothed loss curves and the heatmap-style confusion matrix.
• NumPy: Performs essential array operations, particularly for calculating Intersection over Union (IoU) and filtering statistical outliers in prediction areas.

Section 1: Project Initialization and Dependencies

Purpose: This section establishes the file system structure required for the project and installs the specific external libraries needed for computer vision tasks. It ensures the notebook has access to persistent storage (Google Drive) and the necessary tools for dataset management and inference.

Key Activities:

• Drive Mounting: Connects to Google Drive to create a persistent directory. This is crucial for saving trained weights and logs so they are not lost when the Colab session disconnects.
• Library Installation: Installs roboflow (for downloading datasets), ultralytics (for the YOLO model), and sahi (for Sliced Aided Hyper Inference, used for detecting small objects).
• Credential Management: Securely retrieves the Roboflow API key from Colab’s secrets manager to allow authorized access to the private datasets.

import os

from google.colab import drive

# Mounts Google Drive to allow the Colab notebook to access and save files directly.
drive.mount('/content/drive')

# Defines configuration variables for the main project and the specific model folder.
root_folder = "Mortiscope Models"
model_name = "YOLO26"

# Constructs the full, platform-independent path to the model's project directory.
project_path = os.path.join('/content/drive/MyDrive', root_folder, model_name)
# Constructs the path for a dedicated subdirectory to store model weights.
weights_path = os.path.join(project_path, 'weights')

# Creates the project and weights directories.
os.makedirs(project_path, exist_ok=True)
os.makedirs(weights_path, exist_ok=True)

# Prints the fully constructed paths to the console for user confirmation.
print(f"Project Directory: {project_path}")
print(f"Weight Storage:   {weights_path}")

Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount("/content/drive", force_remount=True).
Project Directory: /content/drive/MyDrive/Mortiscope Models/YOLO26
Weight Storage:   /content/drive/MyDrive/Mortiscope Models/YOLO26/weights

# Executes the shell command to install the specified libraries.
!pip install roboflow ultralytics sahi shapely

# Prints a confirmation message to the console after the installation command completes.
print("Libraries installed successfully.")

Requirement already satisfied: roboflow in /usr/local/lib/python3.12/dist-packages (1.2.12)
Requirement already satisfied: ultralytics in /usr/local/lib/python3.12/dist-packages (8.4.5)
Requirement already satisfied: sahi in /usr/local/lib/python3.12/dist-packages (0.11.36)
Requirement already satisfied: shapely in /usr/local/lib/python3.12/dist-packages (2.1.2)
Requirement already satisfied: certifi in /usr/local/lib/python3.12/dist-packages (from roboflow) (2026.1.4)
Requirement already satisfied: idna==3.7 in /usr/local/lib/python3.12/dist-packages (from roboflow) (3.7)
Requirement already satisfied: cycler in /usr/local/lib/python3.12/dist-packages (from roboflow) (0.12.1)
Requirement already satisfied: kiwisolver>=1.3.1 in /usr/local/lib/python3.12/dist-packages (from roboflow) (1.4.9)
Requirement already satisfied: matplotlib in /usr/local/lib/python3.12/dist-packages (from roboflow) (3.10.0)
Requirement already satisfied: numpy>=1.18.5 in /usr/local/lib/python3.12/dist-packages (from roboflow) (2.0.2)
Requirement already satisfied: opencv-python-headless==4.10.0.84 in /usr/local/lib/python3.12/dist-packages (from roboflow) (4.10.0.84)
Requirement already satisfied: Pillow>=7.1.2 in /usr/local/lib/python3.12/dist-packages (from roboflow) (11.3.0)
Requirement already satisfied: pi-heif<2 in /usr/local/lib/python3.12/dist-packages (from roboflow) (1.1.1)
Requirement already satisfied: pillow-avif-plugin<2 in /usr/local/lib/python3.12/dist-packages (from roboflow) (1.5.2)
Requirement already satisfied: python-dateutil in /usr/local/lib/python3.12/dist-packages (from roboflow) (2.9.0.post0)
Requirement already satisfied: python-dotenv in /usr/local/lib/python3.12/dist-packages (from roboflow) (1.2.1)
Requirement already satisfied: requests in /usr/local/lib/python3.12/dist-packages (from roboflow) (2.32.4)
Requirement already satisfied: six in /usr/local/lib/python3.12/dist-packages (from roboflow) (1.17.0)
Requirement already satisfied: urllib3>=1.26.6 in /usr/local/lib/python3.12/dist-packages (from roboflow) (2.5.0)
Requirement already satisfied: tqdm>=4.41.0 in /usr/local/lib/python3.12/dist-packages (from roboflow) (4.67.1)
Requirement already satisfied: PyYAML>=5.3.1 in /usr/local/lib/python3.12/dist-packages (from roboflow) (6.0.3)
Requirement already satisfied: requests-toolbelt in /usr/local/lib/python3.12/dist-packages (from roboflow) (1.0.0)
Requirement already satisfied: filetype in /usr/local/lib/python3.12/dist-packages (from roboflow) (1.2.0)
Requirement already satisfied: opencv-python>=4.6.0 in /usr/local/lib/python3.12/dist-packages (from ultralytics) (4.11.0.86)
Requirement already satisfied: scipy>=1.4.1 in /usr/local/lib/python3.12/dist-packages (from ultralytics) (1.16.3)
Requirement already satisfied: torch>=1.8.0 in /usr/local/lib/python3.12/dist-packages (from ultralytics) (2.9.0+cu126)
Requirement already satisfied: torchvision>=0.9.0 in /usr/local/lib/python3.12/dist-packages (from ultralytics) (0.24.0+cu126)
Requirement already satisfied: psutil>=5.8.0 in /usr/local/lib/python3.12/dist-packages (from ultralytics) (5.9.5)
Requirement already satisfied: polars>=0.20.0 in /usr/local/lib/python3.12/dist-packages (from ultralytics) (1.31.0)
Requirement already satisfied: ultralytics-thop>=2.0.18 in /usr/local/lib/python3.12/dist-packages (from ultralytics) (2.0.18)
Requirement already satisfied: click in /usr/local/lib/python3.12/dist-packages (from sahi) (8.3.1)
Requirement already satisfied: fire in /usr/local/lib/python3.12/dist-packages (from sahi) (0.7.1)
Requirement already satisfied: pybboxes==0.1.6 in /usr/local/lib/python3.12/dist-packages (from sahi) (0.1.6)
Requirement already satisfied: terminaltables in /usr/local/lib/python3.12/dist-packages (from sahi) (3.1.10)
Requirement already satisfied: contourpy>=1.0.1 in /usr/local/lib/python3.12/dist-packages (from matplotlib->roboflow) (1.3.3)
Requirement already satisfied: fonttools>=4.22.0 in /usr/local/lib/python3.12/dist-packages (from matplotlib->roboflow) (4.61.1)
Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.12/dist-packages (from matplotlib->roboflow) (25.0)
Requirement already satisfied: pyparsing>=2.3.1 in /usr/local/lib/python3.12/dist-packages (from matplotlib->roboflow) (3.3.1)
Requirement already satisfied: charset_normalizer<4,>=2 in /usr/local/lib/python3.12/dist-packages (from requests->roboflow) (3.4.4)
Requirement already satisfied: filelock in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (3.20.2)
Requirement already satisfied: typing-extensions>=4.10.0 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (4.15.0)
Requirement already satisfied: setuptools in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (75.2.0)
Requirement already satisfied: sympy>=1.13.3 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (1.14.0)
Requirement already satisfied: networkx>=2.5.1 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (3.6.1)
Requirement already satisfied: jinja2 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (3.1.6)
Requirement already satisfied: fsspec>=0.8.5 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (2025.3.0)
Requirement already satisfied: nvidia-cuda-nvrtc-cu12==12.6.77 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (12.6.77)
Requirement already satisfied: nvidia-cuda-runtime-cu12==12.6.77 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (12.6.77)
Requirement already satisfied: nvidia-cuda-cupti-cu12==12.6.80 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (12.6.80)
Requirement already satisfied: nvidia-cudnn-cu12==9.10.2.21 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (9.10.2.21)
Requirement already satisfied: nvidia-cublas-cu12==12.6.4.1 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (12.6.4.1)
Requirement already satisfied: nvidia-cufft-cu12==11.3.0.4 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (11.3.0.4)
Requirement already satisfied: nvidia-curand-cu12==10.3.7.77 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (10.3.7.77)
Requirement already satisfied: nvidia-cusolver-cu12==11.7.1.2 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (11.7.1.2)
Requirement already satisfied: nvidia-cusparse-cu12==12.5.4.2 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (12.5.4.2)
Requirement already satisfied: nvidia-cusparselt-cu12==0.7.1 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (0.7.1)
Requirement already satisfied: nvidia-nccl-cu12==2.27.5 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (2.27.5)
Requirement already satisfied: nvidia-nvshmem-cu12==3.3.20 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (3.3.20)
Requirement already satisfied: nvidia-nvtx-cu12==12.6.77 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (12.6.77)
Requirement already satisfied: nvidia-nvjitlink-cu12==12.6.85 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (12.6.85)
Requirement already satisfied: nvidia-cufile-cu12==1.11.1.6 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (1.11.1.6)
Requirement already satisfied: triton==3.5.0 in /usr/local/lib/python3.12/dist-packages (from torch>=1.8.0->ultralytics) (3.5.0)
Requirement already satisfied: termcolor in /usr/local/lib/python3.12/dist-packages (from fire->sahi) (3.3.0)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in /usr/local/lib/python3.12/dist-packages (from sympy>=1.13.3->torch>=1.8.0->ultralytics) (1.3.0)
Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.12/dist-packages (from jinja2->torch>=1.8.0->ultralytics) (3.0.3)
Libraries installed successfully.

from google.colab import userdata
from roboflow import Roboflow

try:
    # Attempts to retrieve the 'ROBOFLOW_API_KEY' from Google Colab's secret manager.
    rf_api_key = userdata.get('ROBOFLOW_API_KEY')
    print("API Key retrieved successfully.")
except Exception as e:
    # Handles potential errors, such as the secret not being defined in the Colab environment.
    print("Error: Could not retrieve key.")
    # Re-raises the exception to halt execution if the key is essential.
    raise e

# Initializes the Roboflow client with the successfully retrieved API key.
rf = Roboflow(api_key=rf_api_key)

API Key retrieved successfully.

Section 2: Dataset Acquisition and Configuration

Purpose: To aggregate data from multiple sources and format it for the YOLO26 training pipeline. Since the dataset is versioned across different projects (representing different resolutions and environmental conditions), this section consolidates them into a single training configuration.

Key Activities:

• Dataset Download: Uses the Roboflow API to download specific versions of the Chrysomya megacephala datasets.
• Dynamic YAML Generation: Instead of manually creating a configuration file, the script iterates through the downloaded folders to verify valid paths. It then programmatically generates a data.yaml file to ensure that the model only attempts to train on data that actually exists locally, preventing “path not found” errors.

# Accesses the specific Roboflow workspace that contains all the project datasets.
workspace = rf.workspace("mortiscope-fvkhd")

# Downloads version 1 of projects using the "yolo26" format.
decomposition_1 = workspace.project("decomposition-high-resolution-300").version(1).download("yolo26")
decomposition_2 = workspace.project("decomposition-high-resolution-250").version(1).download("yolo26")
decomposition_3 = workspace.project("decomposition-high-resolution-200").version(1).download("yolo26")
decomposition_4 = workspace.project("decomposition-standard-resolution-300").version(1).download("yolo26")
decomposition_5 = workspace.project("decomposition-standard-resolution-250").version(1).download("yolo26")
complementary = workspace.project("complementary").version(1).download("yolo26")

# Prints a confirmation message to the console after all download operations have successfully completed.
print("\nAll datasets downloaded successfully.")

loading Roboflow workspace...
loading Roboflow project...
loading Roboflow project...
loading Roboflow project...
loading Roboflow project...
loading Roboflow project...
loading Roboflow project...

All datasets downloaded successfully.

import yaml

# A list containing all the dataset objects that were previously downloaded.
all_datasets = [
    decomposition_1,
    decomposition_2,
    decomposition_3,
    decomposition_4,
    decomposition_5,
    complementary
]

# Initializes lists to store the file paths to the training and validation image folders.
train_paths = []
val_paths = []

print("-"*70)
print("Building Dataset Configuration")

# Iterates through each dataset to locate and collect the paths to its image directories.
for ds in all_datasets:
    # Constructs the expected paths for the 'train' and 'valid' image subdirectories.
    t_path = os.path.join(ds.location, 'train', 'images')
    v_path = os.path.join(ds.location, 'valid', 'images')

    # Verifies that the training directory actually exists before adding it to the list.
    if os.path.exists(t_path):
        train_paths.append(t_path)
        print(f"Added train: {ds.location.split('/')[-1]}")
    else:
        print(f"Skipped train (Empty): {ds.location.split('/')[-1]}")

    # Verifies that the validation directory actually exists before adding it.
    if os.path.exists(v_path):
        val_paths.append(v_path)
        print(f"Added valid: {ds.location.split('/')[-1]}")
    else:
        print(f"Skipped valid (Not found): {ds.location.split('/')[-1]}")

# Defines the master configuration dictionary in the format required by Ultralytics.
data_config = {
    'names': {
        0: 'adult',
        1: 'instar_1',
        2: 'instar_2',
        3: 'instar_3',
        4: 'pupa'
    },
    'nc': 5, # Number of classes.
    'train': train_paths,
    'val': val_paths,
}

# Defines the output path for the YAML file in the current working directory.
yaml_path = os.path.join(os.getcwd(), 'data.yaml')

# Writes the configuration dictionary to the 'data.yaml' file.
with open(yaml_path, 'w') as outfile:
    yaml.dump(data_config, outfile, default_flow_style=False)

# Prints a confirmation summary.
print("\n" + "-"*70)
print(f"Balanced configuration created at: {yaml_path}")
print(f"Total Train Folders: {len(train_paths)}")
print(f"Total Valid Folders: {len(val_paths)}")
print("-"*70)

----------------------------------------------------------------------
Building Dataset Configuration
Added train: Decomposition-(High-Resolution---300%)-1
Added valid: Decomposition-(High-Resolution---300%)-1
Added train: Decomposition-(High-Resolution---250%)-1
Added valid: Decomposition-(High-Resolution---250%)-1
Added train: Decomposition-(High-Resolution---200%)-1
Skipped valid (Not found): Decomposition-(High-Resolution---200%)-1
Added train: Decomposition-(Standard-Resolution---300%)-1
Added valid: Decomposition-(Standard-Resolution---300%)-1
Added train: Decomposition-(Standard-Resolution---250%)-1
Added valid: Decomposition-(Standard-Resolution---250%)-1
Added train: Complementary-1
Added valid: Complementary-1
----------------------------------------------------------------------
Balanced configuration created at: /content/data.yaml
Total Train Folders: 6
Total Valid Folders: 5
----------------------------------------------------------------------

Section 3: Model Initialization and Callback Definition

Purpose: To prepare the architecture and define custom behaviors during the training process. This section ensures that training progress is saved frequently and securely.

Key Activities:

• Custom Callback (on_train_epoch_end): A custom function is defined to run after every training epoch. It copies the model weights (last.pt and best.pt) from the temporary Colab runtime to the persistent Google Drive folder. It also creates a historical archive of weights for every epoch, allowing for granular analysis of model convergence later.
• Model Instantiation: Downloads the base weights for YOLO26l (Large). The large variant is specifically selected for the training in this notebook.

import shutil

# Defines and creates a dedicated directory for storing a snapshot of the model weights at the end of each training epoch.
history_path = os.path.join(weights_path, 'epoch_history')
os.makedirs(history_path, exist_ok=True)


def on_train_epoch_end(trainer):
    """
    A callback function executed at the end of each training epoch.

    This function performs the following actions:
    1. Saves a copy of the latest model weights (`last.pt`) to a historical
       archive, named with the corresponding epoch number.
    2. Updates the primary `last.pt` file in the persistent Google Drive
       weights folder, allowing for training resumption.
    3. Updates the `best.pt` file in the persistent folder whenever the trainer
       identifies a new best-performing model.

    Args:
        trainer: The Ultralytics trainer object, which provides access to the
                 current training state, including epoch number and file paths.
    """
    # Gets the current epoch number.
    current_epoch = trainer.epoch + 1

    # Defines the source paths for the weights generated by the trainer in the temporary, session-specific output directory.
    local_last = os.path.join(trainer.save_dir, 'weights', 'last.pt')
    local_best = os.path.join(trainer.save_dir, 'weights', 'best.pt')

    # Checks for the existence of the latest epoch's weights before proceeding.
    if os.path.exists(local_last):
        # Creates a unique filename for the historical weight file.
        history_filename = f"{current_epoch:03d}_epoch.pt"
        history_dest = os.path.join(history_path, history_filename)

        # Copies the latest weights to the historical archive directory.
        shutil.copy(local_last, history_dest)
        print(f"   History Saved: {history_filename}")

        # Overwrites the main 'last.pt' file in the persistent Google Drive folder.
        resume_dest = os.path.join(weights_path, 'last.pt')
        shutil.copy(local_last, resume_dest)

    # Checks if the trainer has produced a new best-performing model weight file.
    if os.path.exists(local_best):
        # Overwrites the main 'best.pt' file in the persistent Google Drive folder.
        best_dest = os.path.join(weights_path, 'best.pt')
        shutil.copy(local_best, best_dest)

# Prints a confirmation message detailing the callback's configuration and the locations where files will be saved.
print("Callback Defined: ")
print(f"   1. History saved to: {history_path}")
print(f"   2. Resume file active at: {weights_path}/last.pt")

Callback Defined: 
   1. History saved to: /content/drive/MyDrive/Mortiscope Models/YOLO26/weights/epoch_history
   2. Resume file active at: /content/drive/MyDrive/Mortiscope Models/YOLO26/weights/last.pt

import torch
from ultralytics import YOLO

# Defines a list of URLs for various pre-trained YOLO26 model sizes, ranging from nano (n) to extra-large (x).
urls = [
    "https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26n.pt",
    "https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26s.pt",
    "https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26m.pt",
    "https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26l.pt",
    "https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26x.pt"
]

# Selects a specific model variant to serve as the training base.
selected_url = urls[3]
model_filename = os.path.basename(selected_url)

# Prevents re-downloading the model file if it already exists locally.
if not os.path.exists(model_filename):
    print(f"Downloading {model_filename}...")
    # Uses torch.hub to download the file from the specified URL.
    torch.hub.download_url_to_file(selected_url, model_filename)
else:
    print(f"{model_filename} found locally.")

# Instantiates the YOLO model, loading the architecture and pre-trained weights from the downloaded '.pt' file.
model = YOLO(model_filename)

# Registers the custom `on_train_epoch_end` callback function.
model.add_callback("on_train_epoch_end", on_train_epoch_end)

# Prints a summary of the model's architecture.
print("\nModel Architecture & Complexity Base:")
model.info(detailed=True)

yolo26l.pt found locally.

Model Architecture & Complexity Base:
layer                                    name                type  gradient  parameters               shape        mu     sigma
    0                     model.0.conv.weight              Conv2d     False        1728       [64, 3, 3, 3]  -0.00102     0.363        float32
    1                       model.0.bn.weight         BatchNorm2d     False          64                [64]      4.49     0.619        float32
    1                         model.0.bn.bias         BatchNorm2d     False          64                [64]    -0.303      2.16        float32
    2                             model.0.act                SiLU     False           0                  []         -         -              -
    3                     model.1.conv.weight              Conv2d     False       73728     [128, 64, 3, 3] -0.000351    0.0383        float32
    4                       model.1.bn.weight         BatchNorm2d     False         128               [128]      2.75     0.561        float32
    4                         model.1.bn.bias         BatchNorm2d     False         128               [128]     0.822      1.43        float32
    5                 model.2.cv1.conv.weight              Conv2d     False       16384    [128, 128, 1, 1]  -0.00388    0.0602        float32
    6                   model.2.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.14     0.562        float32
    6                     model.2.cv1.bn.bias         BatchNorm2d     False         128               [128]      0.52      1.04        float32
    7                 model.2.cv2.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00268    0.0399        float32
    8                   model.2.cv2.bn.weight         BatchNorm2d     False         256               [256]     0.764     0.194        float32
    8                     model.2.cv2.bn.bias         BatchNorm2d     False         256               [256]    -0.787      0.56        float32
    9             model.2.m.0.cv1.conv.weight              Conv2d     False        2048      [32, 64, 1, 1]  -0.00552    0.0773        float32
   10               model.2.m.0.cv1.bn.weight         BatchNorm2d     False          32                [32]     0.583     0.374        float32
   10                 model.2.m.0.cv1.bn.bias         BatchNorm2d     False          32                [32]     0.588     0.855        float32
   11             model.2.m.0.cv2.conv.weight              Conv2d     False        2048      [32, 64, 1, 1]  -0.00277    0.0808        float32
   12               model.2.m.0.cv2.bn.weight         BatchNorm2d     False          32                [32]     0.979     0.479        float32
   12                 model.2.m.0.cv2.bn.bias         BatchNorm2d     False          32                [32]      1.08     0.853        float32
   13             model.2.m.0.cv3.conv.weight              Conv2d     False        4096      [64, 64, 1, 1]  -0.00283    0.0812        float32
   14               model.2.m.0.cv3.bn.weight         BatchNorm2d     False          64                [64]     0.963      0.29        float32
   14                 model.2.m.0.cv3.bn.bias         BatchNorm2d     False          64                [64]     0.397     0.787        float32
   15         model.2.m.0.m.0.cv1.conv.weight              Conv2d     False        9216      [32, 32, 3, 3] -0.000861    0.0461        float32
   16           model.2.m.0.m.0.cv1.bn.weight         BatchNorm2d     False          32                [32]      1.27     0.436        float32
   16             model.2.m.0.m.0.cv1.bn.bias         BatchNorm2d     False          32                [32]   -0.0399     0.916        float32
   17         model.2.m.0.m.0.cv2.conv.weight              Conv2d     False        9216      [32, 32, 3, 3]  -0.00374    0.0405        float32
   18           model.2.m.0.m.0.cv2.bn.weight         BatchNorm2d     False          32                [32]     0.626     0.319        float32
   18             model.2.m.0.m.0.cv2.bn.bias         BatchNorm2d     False          32                [32]     0.547     0.616        float32
   19         model.2.m.0.m.1.cv1.conv.weight              Conv2d     False        9216      [32, 32, 3, 3] -0.000947    0.0444        float32
   20           model.2.m.0.m.1.cv1.bn.weight         BatchNorm2d     False          32                [32]     0.801     0.263        float32
   20             model.2.m.0.m.1.cv1.bn.bias         BatchNorm2d     False          32                [32]   -0.0561      1.03        float32
   21         model.2.m.0.m.1.cv2.conv.weight              Conv2d     False        9216      [32, 32, 3, 3]  0.000112    0.0421        float32
   22           model.2.m.0.m.1.cv2.bn.weight         BatchNorm2d     False          32                [32]     0.816     0.256        float32
   22             model.2.m.0.m.1.cv2.bn.bias         BatchNorm2d     False          32                [32]     0.907     0.895        float32
   23             model.2.m.1.cv1.conv.weight              Conv2d     False        2048      [32, 64, 1, 1]  -0.00506    0.0774        float32
   24               model.2.m.1.cv1.bn.weight         BatchNorm2d     False          32                [32]     0.317     0.126        float32
   24                 model.2.m.1.cv1.bn.bias         BatchNorm2d     False          32                [32]     0.157     0.466        float32
   25             model.2.m.1.cv2.conv.weight              Conv2d     False        2048      [32, 64, 1, 1]  -0.00646    0.0683        float32
   26               model.2.m.1.cv2.bn.weight         BatchNorm2d     False          32                [32]     0.947     0.382        float32
   26                 model.2.m.1.cv2.bn.bias         BatchNorm2d     False          32                [32]    -0.536     0.524        float32
   27             model.2.m.1.cv3.conv.weight              Conv2d     False        4096      [64, 64, 1, 1]  -0.00174    0.0705        float32
   28               model.2.m.1.cv3.bn.weight         BatchNorm2d     False          64                [64]      1.14     0.176        float32
   28                 model.2.m.1.cv3.bn.bias         BatchNorm2d     False          64                [64]     0.391     0.528        float32
   29         model.2.m.1.m.0.cv1.conv.weight              Conv2d     False        9216      [32, 32, 3, 3]  -0.00178    0.0417        float32
   30           model.2.m.1.m.0.cv1.bn.weight         BatchNorm2d     False          32                [32]     0.693     0.112        float32
   30             model.2.m.1.m.0.cv1.bn.bias         BatchNorm2d     False          32                [32]    -0.163     0.556        float32
   31         model.2.m.1.m.0.cv2.conv.weight              Conv2d     False        9216      [32, 32, 3, 3]  -0.00264    0.0392        float32
   32           model.2.m.1.m.0.cv2.bn.weight         BatchNorm2d     False          32                [32]     0.515     0.143        float32
   32             model.2.m.1.m.0.cv2.bn.bias         BatchNorm2d     False          32                [32]     0.251     0.556        float32
   33         model.2.m.1.m.1.cv1.conv.weight              Conv2d     False        9216      [32, 32, 3, 3]  -0.00224    0.0398        float32
   34           model.2.m.1.m.1.cv1.bn.weight         BatchNorm2d     False          32                [32]     0.597    0.0941        float32
   34             model.2.m.1.m.1.cv1.bn.bias         BatchNorm2d     False          32                [32]    -0.392     0.539        float32
   35         model.2.m.1.m.1.cv2.conv.weight              Conv2d     False        9216      [32, 32, 3, 3] -9.48e-05    0.0436        float32
   36           model.2.m.1.m.1.cv2.bn.weight         BatchNorm2d     False          32                [32]     0.654     0.131        float32
   36             model.2.m.1.m.1.cv2.bn.bias         BatchNorm2d     False          32                [32]       1.2     0.655        float32
   37                     model.3.conv.weight              Conv2d     False      589824    [256, 256, 3, 3] -0.000389    0.0163        float32
   38                       model.3.bn.weight         BatchNorm2d     False         256               [256]     0.643     0.189        float32
   38                         model.3.bn.bias         BatchNorm2d     False         256               [256]     0.132     0.852        float32
   39                 model.4.cv1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00162    0.0386        float32
   40                   model.4.cv1.bn.weight         BatchNorm2d     False         256               [256]      0.64     0.174        float32
   40                     model.4.cv1.bn.bias         BatchNorm2d     False         256               [256]      0.22      0.66        float32
   41                 model.4.cv2.conv.weight              Conv2d     False      262144    [512, 512, 1, 1]  -0.00122    0.0271        float32
   42                   model.4.cv2.bn.weight         BatchNorm2d     False         512               [512]     0.726     0.179        float32
   42                     model.4.cv2.bn.bias         BatchNorm2d     False         512               [512]    -0.817     0.556        float32
   43             model.4.m.0.cv1.conv.weight              Conv2d     False        8192     [64, 128, 1, 1]  -0.00266    0.0522        float32
   44               model.4.m.0.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.394     0.118        float32
   44                 model.4.m.0.cv1.bn.bias         BatchNorm2d     False          64                [64]     0.331      0.44        float32
   45             model.4.m.0.cv2.conv.weight              Conv2d     False        8192     [64, 128, 1, 1]  -0.00147      0.05        float32
   46               model.4.m.0.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.712     0.126        float32
   46                 model.4.m.0.cv2.bn.bias         BatchNorm2d     False          64                [64]     0.253     0.346        float32
   47             model.4.m.0.cv3.conv.weight              Conv2d     False       16384    [128, 128, 1, 1]  -0.00386    0.0497        float32
   48               model.4.m.0.cv3.bn.weight         BatchNorm2d     False         128               [128]     0.755     0.139        float32
   48                 model.4.m.0.cv3.bn.bias         BatchNorm2d     False         128               [128]    -0.178     0.511        float32
   49         model.4.m.0.m.0.cv1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]   -0.0013    0.0285        float32
   50           model.4.m.0.m.0.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.815     0.147        float32
   50             model.4.m.0.m.0.cv1.bn.bias         BatchNorm2d     False          64                [64]    -0.536     0.582        float32
   51         model.4.m.0.m.0.cv2.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]   -0.0015    0.0271        float32
   52           model.4.m.0.m.0.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.578     0.164        float32
   52             model.4.m.0.m.0.cv2.bn.bias         BatchNorm2d     False          64                [64]     0.154      0.42        float32
   53         model.4.m.0.m.1.cv1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00142    0.0274        float32
   54           model.4.m.0.m.1.cv1.bn.weight         BatchNorm2d     False          64                [64]       0.8     0.126        float32
   54             model.4.m.0.m.1.cv1.bn.bias         BatchNorm2d     False          64                [64]    -0.625      0.59        float32
   55         model.4.m.0.m.1.cv2.conv.weight              Conv2d     False       36864      [64, 64, 3, 3] -0.000841    0.0269        float32
   56           model.4.m.0.m.1.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.757     0.189        float32
   56             model.4.m.0.m.1.cv2.bn.bias         BatchNorm2d     False          64                [64]     0.533      0.53        float32
   57             model.4.m.1.cv1.conv.weight              Conv2d     False        8192     [64, 128, 1, 1]  -0.00383    0.0518        float32
   58               model.4.m.1.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.473     0.125        float32
   58                 model.4.m.1.cv1.bn.bias         BatchNorm2d     False          64                [64]   -0.0386      0.48        float32
   59             model.4.m.1.cv2.conv.weight              Conv2d     False        8192     [64, 128, 1, 1]  -0.00297     0.046        float32
   60               model.4.m.1.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.995     0.124        float32
   60                 model.4.m.1.cv2.bn.bias         BatchNorm2d     False          64                [64]    -0.293     0.294        float32
   61             model.4.m.1.cv3.conv.weight              Conv2d     False       16384    [128, 128, 1, 1]  -0.00346    0.0459        float32
   62               model.4.m.1.cv3.bn.weight         BatchNorm2d     False         128               [128]     0.862     0.198        float32
   62                 model.4.m.1.cv3.bn.bias         BatchNorm2d     False         128               [128]    -0.207     0.498        float32
   63         model.4.m.1.m.0.cv1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00133    0.0263        float32
   64           model.4.m.1.m.0.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.801    0.0889        float32
   64             model.4.m.1.m.0.cv1.bn.bias         BatchNorm2d     False          64                [64]    -0.825     0.382        float32
   65         model.4.m.1.m.0.cv2.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00127    0.0255        float32
   66           model.4.m.1.m.0.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.585     0.146        float32
   66             model.4.m.1.m.0.cv2.bn.bias         BatchNorm2d     False          64                [64]   -0.0658     0.384        float32
   67         model.4.m.1.m.1.cv1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00136    0.0265        float32
   68           model.4.m.1.m.1.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.787     0.109        float32
   68             model.4.m.1.m.1.cv1.bn.bias         BatchNorm2d     False          64                [64]     -0.58     0.408        float32
   69         model.4.m.1.m.1.cv2.conv.weight              Conv2d     False       36864      [64, 64, 3, 3] -0.000641    0.0264        float32
   70           model.4.m.1.m.1.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.952     0.195        float32
   70             model.4.m.1.m.1.cv2.bn.bias         BatchNorm2d     False          64                [64]     0.833     0.655        float32
   71                     model.5.conv.weight              Conv2d     False  2.3593e+06    [512, 512, 3, 3] -0.000128    0.0102        float32
   72                       model.5.bn.weight         BatchNorm2d     False         512               [512]     0.785     0.183        float32
   72                         model.5.bn.bias         BatchNorm2d     False         512               [512]    -0.311     0.538        float32
   73                 model.6.cv1.conv.weight              Conv2d     False      262144    [512, 512, 1, 1]   -0.0017    0.0241        float32
   74                   model.6.cv1.bn.weight         BatchNorm2d     False         512               [512]     0.882     0.164        float32
   74                     model.6.cv1.bn.bias         BatchNorm2d     False         512               [512]    -0.312      0.46        float32
   75                 model.6.cv2.conv.weight              Conv2d     False      524288   [512, 1024, 1, 1] -0.000931    0.0198        float32
   76                   model.6.cv2.bn.weight         BatchNorm2d     False         512               [512]     0.811     0.172        float32
   76                     model.6.cv2.bn.bias         BatchNorm2d     False         512               [512]    -0.794     0.615        float32
   77             model.6.m.0.cv1.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00156    0.0348        float32
   78               model.6.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.547     0.132        float32
   78                 model.6.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]     0.121     0.497        float32
   79             model.6.m.0.cv2.conv.weight              Conv2d     False       32768    [128, 256, 1, 1] -0.000514    0.0335        float32
   80               model.6.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.965     0.114        float32
   80                 model.6.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.161     0.267        float32
   81             model.6.m.0.cv3.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00215    0.0321        float32
   82               model.6.m.0.cv3.bn.weight         BatchNorm2d     False         256               [256]     0.902     0.141        float32
   82                 model.6.m.0.cv3.bn.bias         BatchNorm2d     False         256               [256]    -0.418     0.396        float32
   83         model.6.m.0.m.0.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00108    0.0193        float32
   84           model.6.m.0.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.05     0.129        float32
   84             model.6.m.0.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]     -0.64     0.322        float32
   85         model.6.m.0.m.0.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000622     0.018        float32
   86           model.6.m.0.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.828     0.178        float32
   86             model.6.m.0.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.174     0.419        float32
   87         model.6.m.0.m.1.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00114    0.0182        float32
   88           model.6.m.0.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.04     0.131        float32
   88             model.6.m.0.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.651      0.37        float32
   89         model.6.m.0.m.1.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000503    0.0179        float32
   90           model.6.m.0.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.14     0.248        float32
   90             model.6.m.0.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]    0.0419     0.491        float32
   91             model.6.m.1.cv1.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00332    0.0335        float32
   92               model.6.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.683     0.131        float32
   92                 model.6.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]     -0.39      0.45        float32
   93             model.6.m.1.cv2.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00247    0.0318        float32
   94               model.6.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.12    0.0964        float32
   94                 model.6.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.252     0.119        float32
   95             model.6.m.1.cv3.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00183    0.0309        float32
   96               model.6.m.1.cv3.bn.weight         BatchNorm2d     False         256               [256]      1.12     0.164        float32
   96                 model.6.m.1.cv3.bn.bias         BatchNorm2d     False         256               [256]    -0.483     0.293        float32
   97         model.6.m.1.m.0.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000994     0.018        float32
   98           model.6.m.1.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.06      0.12        float32
   98             model.6.m.1.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.698     0.315        float32
   99         model.6.m.1.m.0.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000845    0.0175        float32
  100           model.6.m.1.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.945     0.174        float32
  100             model.6.m.1.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.247     0.323        float32
  101         model.6.m.1.m.1.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00102    0.0176        float32
  102           model.6.m.1.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.06     0.118        float32
  102             model.6.m.1.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.659     0.337        float32
  103         model.6.m.1.m.1.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00055    0.0173        float32
  104           model.6.m.1.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.31     0.195        float32
  104             model.6.m.1.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]    0.0873     0.304        float32
  105                     model.7.conv.weight              Conv2d     False  2.3593e+06    [512, 512, 3, 3] -0.000209   0.00933        float32
  106                       model.7.bn.weight         BatchNorm2d     False         512               [512]     0.864     0.161        float32
  106                         model.7.bn.bias         BatchNorm2d     False         512               [512]    -0.182     0.546        float32
  107                 model.8.cv1.conv.weight              Conv2d     False      262144    [512, 512, 1, 1]  -0.00133    0.0223        float32
  108                   model.8.cv1.bn.weight         BatchNorm2d     False         512               [512]      1.04     0.175        float32
  108                     model.8.cv1.bn.bias         BatchNorm2d     False         512               [512]    0.0174     0.533        float32
  109                 model.8.cv2.conv.weight              Conv2d     False      524288   [512, 1024, 1, 1] -0.000832    0.0179        float32
  110                   model.8.cv2.bn.weight         BatchNorm2d     False         512               [512]      1.06     0.248        float32
  110                     model.8.cv2.bn.bias         BatchNorm2d     False         512               [512]    0.0786     0.527        float32
  111             model.8.m.0.cv1.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]   -0.0026    0.0312        float32
  112               model.8.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.505     0.126        float32
  112                 model.8.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.226     0.473        float32
  113             model.8.m.0.cv2.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00128    0.0295        float32
  114               model.8.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.11     0.152        float32
  114                 model.8.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.194      0.12        float32
  115             model.8.m.0.cv3.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00161    0.0292        float32
  116               model.8.m.0.cv3.bn.weight         BatchNorm2d     False         256               [256]     0.921     0.202        float32
  116                 model.8.m.0.cv3.bn.bias         BatchNorm2d     False         256               [256]     -0.24     0.328        float32
  117         model.8.m.0.m.0.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00063    0.0176        float32
  118           model.8.m.0.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.07     0.155        float32
  118             model.8.m.0.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.211     0.413        float32
  119         model.8.m.0.m.0.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000458    0.0169        float32
  120           model.8.m.0.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.11     0.213        float32
  120             model.8.m.0.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]   -0.0798     0.359        float32
  121         model.8.m.0.m.1.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000945    0.0169        float32
  122           model.8.m.0.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.07     0.154        float32
  122             model.8.m.0.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.202     0.333        float32
  123         model.8.m.0.m.1.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000652    0.0163        float32
  124           model.8.m.0.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.36     0.322        float32
  124             model.8.m.0.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]     0.115     0.269        float32
  125             model.8.m.1.cv1.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]   -0.0021    0.0294        float32
  126               model.8.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.421      0.15        float32
  126                 model.8.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.249     0.368        float32
  127             model.8.m.1.cv2.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00187    0.0293        float32
  128               model.8.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.11    0.0951        float32
  128                 model.8.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]   -0.0857    0.0932        float32
  129             model.8.m.1.cv3.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00169     0.027        float32
  130               model.8.m.1.cv3.bn.weight         BatchNorm2d     False         256               [256]     0.995     0.224        float32
  130                 model.8.m.1.cv3.bn.bias         BatchNorm2d     False         256               [256]    -0.128     0.256        float32
  131         model.8.m.1.m.0.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000462    0.0164        float32
  132           model.8.m.1.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.07     0.254        float32
  132             model.8.m.1.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.158     0.255        float32
  133         model.8.m.1.m.0.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00058     0.016        float32
  134           model.8.m.1.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.946      0.21        float32
  134             model.8.m.1.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.295     0.224        float32
  135         model.8.m.1.m.1.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000879    0.0161        float32
  136           model.8.m.1.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.08     0.176        float32
  136             model.8.m.1.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.268     0.284        float32
  137         model.8.m.1.m.1.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000686    0.0152        float32
  138           model.8.m.1.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]       1.5     0.356        float32
  138             model.8.m.1.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]   -0.0354     0.182        float32
  139                 model.9.cv1.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  -0.00137    0.0235        float32
  140                   model.9.cv1.bn.weight         BatchNorm2d     False         256               [256]     0.972     0.268        float32
  140                     model.9.cv1.bn.bias         BatchNorm2d     False         256               [256] -1.94e-06  2.94e-05        float32
  141                         model.9.cv1.act            Identity     False           0                  []         -         -              -
  142                 model.9.cv2.conv.weight              Conv2d     False      524288   [512, 1024, 1, 1] -0.000142    0.0174        float32
  143                   model.9.cv2.bn.weight         BatchNorm2d     False         512               [512]      1.05     0.384        float32
  143                     model.9.cv2.bn.bias         BatchNorm2d     False         512               [512]    -0.447      0.34        float32
  144                               model.9.m           MaxPool2d     False           0                  []         -         -              -
  145                model.10.cv1.conv.weight              Conv2d     False      262144    [512, 512, 1, 1]  -0.00077    0.0241        float32
  146                  model.10.cv1.bn.weight         BatchNorm2d     False         512               [512]      1.19     0.375        float32
  146                    model.10.cv1.bn.bias         BatchNorm2d     False         512               [512]     0.293     0.631        float32
  147                model.10.cv2.conv.weight              Conv2d     False      262144    [512, 512, 1, 1] -0.000501    0.0246        float32
  148                  model.10.cv2.bn.weight         BatchNorm2d     False         512               [512]     0.866     0.238        float32
  148                    model.10.cv2.bn.bias         BatchNorm2d     False         512               [512]    -0.542     0.514        float32
  149       model.10.m.0.attn.qkv.conv.weight              Conv2d     False      131072    [512, 256, 1, 1]  1.74e-05    0.0278        float32
  150         model.10.m.0.attn.qkv.bn.weight         BatchNorm2d     False         512               [512]      1.31      0.32        float32
  150           model.10.m.0.attn.qkv.bn.bias         BatchNorm2d     False         512               [512]  -0.00236     0.268        float32
  151               model.10.m.0.attn.qkv.act            Identity     False           0                  []         -         -              -
  152      model.10.m.0.attn.proj.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  5.18e-05    0.0351        float32
  153        model.10.m.0.attn.proj.bn.weight         BatchNorm2d     False         256               [256]     0.672     0.296        float32
  153          model.10.m.0.attn.proj.bn.bias         BatchNorm2d     False         256               [256]  4.23e-06  4.32e-05        float32
  154              model.10.m.0.attn.proj.act            Identity     False           0                  []         -         -              -
  155        model.10.m.0.attn.pe.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]   -0.0195    0.0573        float32
  156          model.10.m.0.attn.pe.bn.weight         BatchNorm2d     False         256               [256]     0.636     0.264        float32
  156            model.10.m.0.attn.pe.bn.bias         BatchNorm2d     False         256               [256] -1.11e-06  3.64e-05        float32
  157                model.10.m.0.attn.pe.act            Identity     False           0                  []         -         -              -
  158          model.10.m.0.ffn.0.conv.weight              Conv2d     False      131072    [512, 256, 1, 1]  -0.00144    0.0263        float32
  159            model.10.m.0.ffn.0.bn.weight         BatchNorm2d     False         512               [512]      1.18     0.214        float32
  159              model.10.m.0.ffn.0.bn.bias         BatchNorm2d     False         512               [512]    -0.693     0.244        float32
  160          model.10.m.0.ffn.1.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  0.000477    0.0258        float32
  161            model.10.m.0.ffn.1.bn.weight         BatchNorm2d     False         256               [256]     0.716      0.25        float32
  161              model.10.m.0.ffn.1.bn.bias         BatchNorm2d     False         256               [256]  2.93e-06  4.22e-05        float32
  162                  model.10.m.0.ffn.1.act            Identity     False           0                  []         -         -              -
  163       model.10.m.1.attn.qkv.conv.weight              Conv2d     False      131072    [512, 256, 1, 1] -0.000106    0.0262        float32
  164         model.10.m.1.attn.qkv.bn.weight         BatchNorm2d     False         512               [512]      1.11     0.365        float32
  164           model.10.m.1.attn.qkv.bn.bias         BatchNorm2d     False         512               [512]    0.0188     0.269        float32
  165               model.10.m.1.attn.qkv.act            Identity     False           0                  []         -         -              -
  166      model.10.m.1.attn.proj.conv.weight              Conv2d     False       65536    [256, 256, 1, 1] -5.31e-05    0.0335        float32
  167        model.10.m.1.attn.proj.bn.weight         BatchNorm2d     False         256               [256]     0.925     0.214        float32
  167          model.10.m.1.attn.proj.bn.bias         BatchNorm2d     False         256               [256] -1.48e-06  2.13e-05        float32
  168              model.10.m.1.attn.proj.act            Identity     False           0                  []         -         -              -
  169        model.10.m.1.attn.pe.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]   -0.0173    0.0502        float32
  170          model.10.m.1.attn.pe.bn.weight         BatchNorm2d     False         256               [256]     0.689      0.27        float32
  170            model.10.m.1.attn.pe.bn.bias         BatchNorm2d     False         256               [256] -2.29e-07  3.41e-05        float32
  171                model.10.m.1.attn.pe.act            Identity     False           0                  []         -         -              -
  172          model.10.m.1.ffn.0.conv.weight              Conv2d     False      131072    [512, 256, 1, 1] -0.000753    0.0265        float32
  173            model.10.m.1.ffn.0.bn.weight         BatchNorm2d     False         512               [512]      1.19     0.165        float32
  173              model.10.m.1.ffn.0.bn.bias         BatchNorm2d     False         512               [512]    -0.599     0.188        float32
  174          model.10.m.1.ffn.1.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  0.000107    0.0261        float32
  175            model.10.m.1.ffn.1.bn.weight         BatchNorm2d     False         256               [256]     0.887     0.244        float32
  175              model.10.m.1.ffn.1.bn.bias         BatchNorm2d     False         256               [256] -1.54e-06  1.99e-05        float32
  176                  model.10.m.1.ffn.1.act            Identity     False           0                  []         -         -              -
  177                                model.11            Upsample     False           0                  []         -         -              -
  178                                model.12              Concat     False           0                  []         -         -              -
  179                model.13.cv1.conv.weight              Conv2d     False      524288   [512, 1024, 1, 1] -0.000718    0.0203        float32
  180                  model.13.cv1.bn.weight         BatchNorm2d     False         512               [512]     0.919     0.205        float32
  180                    model.13.cv1.bn.bias         BatchNorm2d     False         512               [512]    -0.321      0.59        float32
  181                model.13.cv2.conv.weight              Conv2d     False      524288   [512, 1024, 1, 1]  -0.00118    0.0204        float32
  182                  model.13.cv2.bn.weight         BatchNorm2d     False         512               [512]     0.706     0.155        float32
  182                    model.13.cv2.bn.bias         BatchNorm2d     False         512               [512]     -0.63      0.47        float32
  183            model.13.m.0.cv1.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00185    0.0342        float32
  184              model.13.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.695     0.175        float32
  184                model.13.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.176     0.627        float32
  185            model.13.m.0.cv2.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00281    0.0342        float32
  186              model.13.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]       1.3     0.142        float32
  186                model.13.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.277      0.32        float32
  187            model.13.m.0.cv3.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00213     0.032        float32
  188              model.13.m.0.cv3.bn.weight         BatchNorm2d     False         256               [256]     0.892     0.135        float32
  188                model.13.m.0.cv3.bn.bias         BatchNorm2d     False         256               [256]    -0.561      0.48        float32
  189        model.13.m.0.m.0.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000927    0.0181        float32
  190          model.13.m.0.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.08     0.139        float32
  190            model.13.m.0.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.499     0.307        float32
  191        model.13.m.0.m.0.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000722    0.0175        float32
  192          model.13.m.0.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.833      0.16        float32
  192            model.13.m.0.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.239     0.337        float32
  193        model.13.m.0.m.1.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000907    0.0179        float32
  194          model.13.m.0.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.06     0.152        float32
  194            model.13.m.0.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.506     0.268        float32
  195        model.13.m.0.m.1.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000495    0.0173        float32
  196          model.13.m.0.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.989     0.185        float32
  196            model.13.m.0.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]   -0.0319     0.297        float32
  197            model.13.m.1.cv1.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00353    0.0338        float32
  198              model.13.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.705     0.109        float32
  198                model.13.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]     -0.45      0.41        float32
  199            model.13.m.1.cv2.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00254    0.0323        float32
  200              model.13.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.24     0.106        float32
  200                model.13.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.331     0.133        float32
  201            model.13.m.1.cv3.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00291    0.0307        float32
  202              model.13.m.1.cv3.bn.weight         BatchNorm2d     False         256               [256]      1.02     0.128        float32
  202                model.13.m.1.cv3.bn.bias         BatchNorm2d     False         256               [256]    -0.702     0.305        float32
  203        model.13.m.1.m.0.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00123    0.0186        float32
  204          model.13.m.1.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.977     0.109        float32
  204            model.13.m.1.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.798     0.299        float32
  205        model.13.m.1.m.0.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000718    0.0172        float32
  206          model.13.m.1.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.876     0.106        float32
  206            model.13.m.1.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]     -0.22     0.257        float32
  207        model.13.m.1.m.1.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]   -0.0013    0.0176        float32
  208          model.13.m.1.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.982     0.111        float32
  208            model.13.m.1.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.606     0.251        float32
  209        model.13.m.1.m.1.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000526    0.0173        float32
  210          model.13.m.1.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.31     0.153        float32
  210            model.13.m.1.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]    0.0289     0.261        float32
  211                                model.14            Upsample     False           0                  []         -         -              -
  212                                model.15              Concat     False           0                  []         -         -              -
  213                model.16.cv1.conv.weight              Conv2d     False      262144   [256, 1024, 1, 1]  -0.00075    0.0222        float32
  214                  model.16.cv1.bn.weight         BatchNorm2d     False         256               [256]     0.555     0.169        float32
  214                    model.16.cv1.bn.bias         BatchNorm2d     False         256               [256]     0.015     0.793        float32
  215                model.16.cv2.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  -0.00108    0.0285        float32
  216                  model.16.cv2.bn.weight         BatchNorm2d     False         256               [256]     0.701     0.398        float32
  216                    model.16.cv2.bn.bias         BatchNorm2d     False         256               [256]    -0.353     0.746        float32
  217            model.16.m.0.cv1.conv.weight              Conv2d     False        8192     [64, 128, 1, 1]  -0.00331    0.0555        float32
  218              model.16.m.0.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.312     0.106        float32
  218                model.16.m.0.cv1.bn.bias         BatchNorm2d     False          64                [64]     0.245     0.491        float32
  219            model.16.m.0.cv2.conv.weight              Conv2d     False        8192     [64, 128, 1, 1] -0.000184    0.0497        float32
  220              model.16.m.0.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.818      0.12        float32
  220                model.16.m.0.cv2.bn.bias         BatchNorm2d     False          64                [64]     0.529     0.481        float32
  221            model.16.m.0.cv3.conv.weight              Conv2d     False       16384    [128, 128, 1, 1]  -0.00417    0.0505        float32
  222              model.16.m.0.cv3.bn.weight         BatchNorm2d     False         128               [128]     0.715      0.11        float32
  222                model.16.m.0.cv3.bn.bias         BatchNorm2d     False         128               [128]    -0.434     0.413        float32
  223        model.16.m.0.m.0.cv1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00128     0.028        float32
  224          model.16.m.0.m.0.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.833    0.0766        float32
  224            model.16.m.0.m.0.cv1.bn.bias         BatchNorm2d     False          64                [64]    -0.564     0.502        float32
  225        model.16.m.0.m.0.cv2.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00182    0.0275        float32
  226          model.16.m.0.m.0.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.638     0.158        float32
  226            model.16.m.0.m.0.cv2.bn.bias         BatchNorm2d     False          64                [64]    -0.433     0.468        float32
  227        model.16.m.0.m.1.cv1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00161    0.0275        float32
  228          model.16.m.0.m.1.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.821     0.111        float32
  228            model.16.m.0.m.1.cv1.bn.bias         BatchNorm2d     False          64                [64]    -0.594     0.513        float32
  229        model.16.m.0.m.1.cv2.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00103    0.0274        float32
  230          model.16.m.0.m.1.cv2.bn.weight         BatchNorm2d     False          64                [64]      1.08     0.241        float32
  230            model.16.m.0.m.1.cv2.bn.bias         BatchNorm2d     False          64                [64]     0.536     0.506        float32
  231            model.16.m.1.cv1.conv.weight              Conv2d     False        8192     [64, 128, 1, 1]  -0.00352    0.0529        float32
  232              model.16.m.1.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.325     0.122        float32
  232                model.16.m.1.cv1.bn.bias         BatchNorm2d     False          64                [64]    -0.165     0.543        float32
  233            model.16.m.1.cv2.conv.weight              Conv2d     False        8192     [64, 128, 1, 1]   -0.0033    0.0461        float32
  234              model.16.m.1.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.878     0.204        float32
  234                model.16.m.1.cv2.bn.bias         BatchNorm2d     False          64                [64]    -0.186     0.319        float32
  235            model.16.m.1.cv3.conv.weight              Conv2d     False       16384    [128, 128, 1, 1]  -0.00304    0.0485        float32
  236              model.16.m.1.cv3.bn.weight         BatchNorm2d     False         128               [128]      1.19     0.706        float32
  236                model.16.m.1.cv3.bn.bias         BatchNorm2d     False         128               [128]    -0.219     0.698        float32
  237        model.16.m.1.m.0.cv1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00121    0.0281        float32
  238          model.16.m.1.m.0.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.803     0.138        float32
  238            model.16.m.1.m.0.cv1.bn.bias         BatchNorm2d     False          64                [64]    -0.513     0.479        float32
  239        model.16.m.1.m.0.cv2.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00125    0.0277        float32
  240          model.16.m.1.m.0.cv2.bn.weight         BatchNorm2d     False          64                [64]     0.804     0.183        float32
  240            model.16.m.1.m.0.cv2.bn.bias         BatchNorm2d     False          64                [64]    -0.326     0.541        float32
  241        model.16.m.1.m.1.cv1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00148    0.0279        float32
  242          model.16.m.1.m.1.cv1.bn.weight         BatchNorm2d     False          64                [64]     0.833     0.238        float32
  242            model.16.m.1.m.1.cv1.bn.bias         BatchNorm2d     False          64                [64]    -0.695     0.796        float32
  243        model.16.m.1.m.1.cv2.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00057    0.0272        float32
  244          model.16.m.1.m.1.cv2.bn.weight         BatchNorm2d     False          64                [64]      1.54     0.771        float32
  244            model.16.m.1.m.1.cv2.bn.bias         BatchNorm2d     False          64                [64]     0.369     0.777        float32
  245                    model.17.conv.weight              Conv2d     False      589824    [256, 256, 3, 3] -0.000196    0.0131        float32
  246                      model.17.bn.weight         BatchNorm2d     False         256               [256]     0.741     0.121        float32
  246                        model.17.bn.bias         BatchNorm2d     False         256               [256]     -0.33      0.31        float32
  247                                model.18              Concat     False           0                  []         -         -              -
  248                model.19.cv1.conv.weight              Conv2d     False      393216    [512, 768, 1, 1]  -0.00073    0.0209        float32
  249                  model.19.cv1.bn.weight         BatchNorm2d     False         512               [512]     0.928      0.15        float32
  249                    model.19.cv1.bn.bias         BatchNorm2d     False         512               [512]    -0.338     0.387        float32
  250                model.19.cv2.conv.weight              Conv2d     False      524288   [512, 1024, 1, 1] -0.000738    0.0183        float32
  251                  model.19.cv2.bn.weight         BatchNorm2d     False         512               [512]     0.926     0.349        float32
  251                    model.19.cv2.bn.bias         BatchNorm2d     False         512               [512]    -0.326     0.418        float32
  252            model.19.m.0.cv1.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00237    0.0336        float32
  253              model.19.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.453      0.11        float32
  253                model.19.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.189     0.439        float32
  254            model.19.m.0.cv2.conv.weight              Conv2d     False       32768    [128, 256, 1, 1] -0.000506    0.0304        float32
  255              model.19.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.21     0.191        float32
  255                model.19.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    0.0489     0.244        float32
  256            model.19.m.0.cv3.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00203    0.0311        float32
  257              model.19.m.0.cv3.bn.weight         BatchNorm2d     False         256               [256]     0.869     0.169        float32
  257                model.19.m.0.cv3.bn.bias         BatchNorm2d     False         256               [256]    -0.343     0.325        float32
  258        model.19.m.0.m.0.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00101    0.0186        float32
  259          model.19.m.0.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.02    0.0993        float32
  259            model.19.m.0.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.717     0.331        float32
  260        model.19.m.0.m.0.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000861    0.0177        float32
  261          model.19.m.0.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.781    0.0941        float32
  261            model.19.m.0.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.607     0.333        float32
  262        model.19.m.0.m.1.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000883    0.0181        float32
  263          model.19.m.0.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.994    0.0953        float32
  263            model.19.m.0.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.648     0.331        float32
  264        model.19.m.0.m.1.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000622    0.0177        float32
  265          model.19.m.0.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.38     0.179        float32
  265            model.19.m.0.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]   -0.0679     0.263        float32
  266            model.19.m.1.cv1.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00334    0.0329        float32
  267              model.19.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.412     0.128        float32
  267                model.19.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.309     0.434        float32
  268            model.19.m.1.cv2.conv.weight              Conv2d     False       32768    [128, 256, 1, 1]  -0.00236    0.0308        float32
  269              model.19.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.986     0.111        float32
  269                model.19.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.176     0.126        float32
  270            model.19.m.1.cv3.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00182    0.0304        float32
  271              model.19.m.1.cv3.bn.weight         BatchNorm2d     False         256               [256]      1.41      0.46        float32
  271                model.19.m.1.cv3.bn.bias         BatchNorm2d     False         256               [256]    -0.201     0.304        float32
  272        model.19.m.1.m.0.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000752    0.0182        float32
  273          model.19.m.1.m.0.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.971      0.16        float32
  273            model.19.m.1.m.0.cv1.bn.bias         BatchNorm2d     False         128               [128]     -0.73     0.353        float32
  274        model.19.m.1.m.0.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000812    0.0175        float32
  275          model.19.m.1.m.0.cv2.bn.weight         BatchNorm2d     False         128               [128]     0.795     0.204        float32
  275            model.19.m.1.m.0.cv2.bn.bias         BatchNorm2d     False         128               [128]     -0.59       0.4        float32
  276        model.19.m.1.m.1.cv1.conv.weight              Conv2d     False      147456    [128, 128, 3, 3] -0.000867    0.0176        float32
  277          model.19.m.1.m.1.cv1.bn.weight         BatchNorm2d     False         128               [128]     0.909     0.223        float32
  277            model.19.m.1.m.1.cv1.bn.bias         BatchNorm2d     False         128               [128]     -0.77     0.459        float32
  278        model.19.m.1.m.1.cv2.conv.weight              Conv2d     False      147456    [128, 128, 3, 3]  -0.00031    0.0173        float32
  279          model.19.m.1.m.1.cv2.bn.weight         BatchNorm2d     False         128               [128]      1.57     0.471        float32
  279            model.19.m.1.m.1.cv2.bn.bias         BatchNorm2d     False         128               [128]    -0.144     0.374        float32
  280                    model.20.conv.weight              Conv2d     False  2.3593e+06    [512, 512, 3, 3] -0.000151   0.00816        float32
  281                      model.20.bn.weight         BatchNorm2d     False         512               [512]      1.15     0.134        float32
  281                        model.20.bn.bias         BatchNorm2d     False         512               [512]     -0.17     0.114        float32
  282                                model.21              Concat     False           0                  []         -         -              -
  283                model.22.cv1.conv.weight              Conv2d     False      524288   [512, 1024, 1, 1] -0.000531    0.0173        float32
  284                  model.22.cv1.bn.weight         BatchNorm2d     False         512               [512]      1.01      0.26        float32
  284                    model.22.cv1.bn.bias         BatchNorm2d     False         512               [512]    -0.194     0.218        float32
  285                model.22.cv2.conv.weight              Conv2d     False      393216    [512, 768, 1, 1] -0.000252    0.0205        float32
  286                  model.22.cv2.bn.weight         BatchNorm2d     False         512               [512]      1.09     0.319        float32
  286                    model.22.cv2.bn.bias         BatchNorm2d     False         512               [512]    -0.247     0.218        float32
  287          model.22.m.0.0.cv1.conv.weight              Conv2d     False      294912    [128, 256, 3, 3] -0.000529    0.0121        float32
  288            model.22.m.0.0.cv1.bn.weight         BatchNorm2d     False         128               [128]      1.14     0.177        float32
  288              model.22.m.0.0.cv1.bn.bias         BatchNorm2d     False         128               [128]    -0.422     0.321        float32
  289          model.22.m.0.0.cv2.conv.weight              Conv2d     False      294912    [256, 128, 3, 3]  -0.00102    0.0158        float32
  290            model.22.m.0.0.cv2.bn.weight         BatchNorm2d     False         256               [256]      1.17     0.232        float32
  290              model.22.m.0.0.cv2.bn.bias         BatchNorm2d     False         256               [256]    -0.163     0.243        float32
  291     model.22.m.0.1.attn.qkv.conv.weight              Conv2d     False      131072    [512, 256, 1, 1] -3.16e-06    0.0221        float32
  292       model.22.m.0.1.attn.qkv.bn.weight         BatchNorm2d     False         512               [512]      1.16     0.186        float32
  292         model.22.m.0.1.attn.qkv.bn.bias         BatchNorm2d     False         512               [512]  -0.00173     0.166        float32
  293             model.22.m.0.1.attn.qkv.act            Identity     False           0                  []         -         -              -
  294    model.22.m.0.1.attn.proj.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  0.000101    0.0285        float32
  295      model.22.m.0.1.attn.proj.bn.weight         BatchNorm2d     False         256               [256]     0.886    0.0629        float32
  295        model.22.m.0.1.attn.proj.bn.bias         BatchNorm2d     False         256               [256] -8.23e-07  6.38e-06        float32
  296            model.22.m.0.1.attn.proj.act            Identity     False           0                  []         -         -              -
  297      model.22.m.0.1.attn.pe.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]   -0.0108    0.0341        float32
  298        model.22.m.0.1.attn.pe.bn.weight         BatchNorm2d     False         256               [256]      0.85     0.141        float32
  298          model.22.m.0.1.attn.pe.bn.bias         BatchNorm2d     False         256               [256] -1.84e-07  4.45e-06        float32
  299              model.22.m.0.1.attn.pe.act            Identity     False           0                  []         -         -              -
  300        model.22.m.0.1.ffn.0.conv.weight              Conv2d     False      131072    [512, 256, 1, 1] -0.000846    0.0218        float32
  301          model.22.m.0.1.ffn.0.bn.weight         BatchNorm2d     False         512               [512]      1.12     0.118        float32
  301            model.22.m.0.1.ffn.0.bn.bias         BatchNorm2d     False         512               [512]    -0.198    0.0897        float32
  302        model.22.m.0.1.ffn.1.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  0.000234    0.0221        float32
  303          model.22.m.0.1.ffn.1.bn.weight         BatchNorm2d     False         256               [256]      1.01     0.053        float32
  303            model.22.m.0.1.ffn.1.bn.bias         BatchNorm2d     False         256               [256] -5.59e-07  4.51e-06        float32
  304                model.22.m.0.1.ffn.1.act            Identity     False           0                  []         -         -              -
  305            model.23.cv2.0.0.conv.weight              Conv2d     False      147456     [64, 256, 3, 3] -0.000225    0.0141        float32
  306              model.23.cv2.0.0.bn.weight         BatchNorm2d     False          64                [64]       1.4     0.776        float32
  306                model.23.cv2.0.0.bn.bias         BatchNorm2d     False          64                [64]     0.393      1.15        float32
  307            model.23.cv2.0.1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00124    0.0318        float32
  308              model.23.cv2.0.1.bn.weight         BatchNorm2d     False          64                [64]     0.978     0.728        float32
  308                model.23.cv2.0.1.bn.bias         BatchNorm2d     False          64                [64]     -0.19     0.682        float32
  309                 model.23.cv2.0.2.weight              Conv2d     False         256       [4, 64, 1, 1]    0.0452     0.143        float32
  309                   model.23.cv2.0.2.bias              Conv2d     False           4                 [4]      1.16     0.118        float32
  310            model.23.cv2.1.0.conv.weight              Conv2d     False      294912     [64, 512, 3, 3] -0.000155   0.00915        float32
  311              model.23.cv2.1.0.bn.weight         BatchNorm2d     False          64                [64]      1.36     0.495        float32
  311                model.23.cv2.1.0.bn.bias         BatchNorm2d     False          64                [64]    0.0274     0.705        float32
  312            model.23.cv2.1.1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3] -0.000951    0.0293        float32
  313              model.23.cv2.1.1.bn.weight         BatchNorm2d     False          64                [64]     0.949     0.822        float32
  313                model.23.cv2.1.1.bn.bias         BatchNorm2d     False          64                [64]    0.0439     0.522        float32
  314                 model.23.cv2.1.2.weight              Conv2d     False         256       [4, 64, 1, 1]    0.0583     0.147        float32
  314                   model.23.cv2.1.2.bias              Conv2d     False           4                 [4]      1.35     0.176        float32
  315            model.23.cv2.2.0.conv.weight              Conv2d     False      294912     [64, 512, 3, 3] -0.000192   0.00867        float32
  316              model.23.cv2.2.0.bn.weight         BatchNorm2d     False          64                [64]       1.5     0.654        float32
  316                model.23.cv2.2.0.bn.bias         BatchNorm2d     False          64                [64]    0.0634     0.531        float32
  317            model.23.cv2.2.1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00108    0.0271        float32
  318              model.23.cv2.2.1.bn.weight         BatchNorm2d     False          64                [64]      1.11     0.876        float32
  318                model.23.cv2.2.1.bn.bias         BatchNorm2d     False          64                [64]     0.115     0.605        float32
  319                 model.23.cv2.2.2.weight              Conv2d     False         256       [4, 64, 1, 1]    0.0518     0.147        float32
  319                   model.23.cv2.2.2.bias              Conv2d     False           4                 [4]      1.19    0.0904        float32
  320          model.23.cv3.0.0.0.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]    0.0117    0.0417        float32
  321            model.23.cv3.0.0.0.bn.weight         BatchNorm2d     False         256               [256]     0.632     0.285        float32
  321              model.23.cv3.0.0.0.bn.bias         BatchNorm2d     False         256               [256]     0.599     0.633        float32
  322          model.23.cv3.0.0.1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00159    0.0421        float32
  323            model.23.cv3.0.0.1.bn.weight         BatchNorm2d     False         256               [256]     0.783     0.227        float32
  323              model.23.cv3.0.0.1.bn.bias         BatchNorm2d     False         256               [256]    0.0345     0.438        float32
  324          model.23.cv3.0.1.0.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]   0.00622    0.0496        float32
  325            model.23.cv3.0.1.0.bn.weight         BatchNorm2d     False         256               [256]     0.728     0.365        float32
  325              model.23.cv3.0.1.0.bn.bias         BatchNorm2d     False         256               [256]     0.392     0.568        float32
  326          model.23.cv3.0.1.1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1] -0.000521    0.0407        float32
  327            model.23.cv3.0.1.1.bn.weight         BatchNorm2d     False         256               [256]      1.78     0.486        float32
  327              model.23.cv3.0.1.1.bn.bias         BatchNorm2d     False         256               [256]     0.623     0.748        float32
  328                 model.23.cv3.0.2.weight              Conv2d     False       20480     [80, 256, 1, 1]  -0.00905    0.0546        float32
  328                   model.23.cv3.0.2.bias              Conv2d     False          80                [80]     -11.5     0.284        float32
  329          model.23.cv3.1.0.0.conv.weight              Conv2d     False        4608      [512, 1, 3, 3]   0.00679    0.0298        float32
  330            model.23.cv3.1.0.0.bn.weight         BatchNorm2d     False         512               [512]     0.928     0.255        float32
  330              model.23.cv3.1.0.0.bn.bias         BatchNorm2d     False         512               [512]     0.141      0.29        float32
  331          model.23.cv3.1.0.1.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  -0.00155    0.0321        float32
  332            model.23.cv3.1.0.1.bn.weight         BatchNorm2d     False         256               [256]     0.913     0.176        float32
  332              model.23.cv3.1.0.1.bn.bias         BatchNorm2d     False         256               [256]     0.139     0.258        float32
  333          model.23.cv3.1.1.0.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]   0.00344    0.0418        float32
  334            model.23.cv3.1.1.0.bn.weight         BatchNorm2d     False         256               [256]     0.815     0.427        float32
  334              model.23.cv3.1.1.0.bn.bias         BatchNorm2d     False         256               [256]     0.292     0.369        float32
  335          model.23.cv3.1.1.1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00156    0.0383        float32
  336            model.23.cv3.1.1.1.bn.weight         BatchNorm2d     False         256               [256]      1.65     0.394        float32
  336              model.23.cv3.1.1.1.bn.bias         BatchNorm2d     False         256               [256]     0.674      0.87        float32
  337                 model.23.cv3.1.2.weight              Conv2d     False       20480     [80, 256, 1, 1]  -0.00754    0.0542        float32
  337                   model.23.cv3.1.2.bias              Conv2d     False          80                [80]       -10     0.319        float32
  338          model.23.cv3.2.0.0.conv.weight              Conv2d     False        4608      [512, 1, 3, 3]   0.00519    0.0305        float32
  339            model.23.cv3.2.0.0.bn.weight         BatchNorm2d     False         512               [512]     0.988     0.183        float32
  339              model.23.cv3.2.0.0.bn.bias         BatchNorm2d     False         512               [512]    0.0181     0.163        float32
  340          model.23.cv3.2.0.1.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  -0.00177     0.033        float32
  341            model.23.cv3.2.0.1.bn.weight         BatchNorm2d     False         256               [256]      0.97    0.0995        float32
  341              model.23.cv3.2.0.1.bn.bias         BatchNorm2d     False         256               [256]     0.108     0.235        float32
  342          model.23.cv3.2.1.0.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]  0.000336    0.0417        float32
  343            model.23.cv3.2.1.0.bn.weight         BatchNorm2d     False         256               [256]     0.893     0.343        float32
  343              model.23.cv3.2.1.0.bn.bias         BatchNorm2d     False         256               [256]     0.147     0.301        float32
  344          model.23.cv3.2.1.1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00156     0.038        float32
  345            model.23.cv3.2.1.1.bn.weight         BatchNorm2d     False         256               [256]      1.43     0.498        float32
  345              model.23.cv3.2.1.1.bn.bias         BatchNorm2d     False         256               [256]     0.535     0.677        float32
  346                 model.23.cv3.2.2.weight              Conv2d     False       20480     [80, 256, 1, 1]  -0.00884    0.0627        float32
  346                   model.23.cv3.2.2.bias              Conv2d     False          80                [80]     -8.76     0.285        float32
  347                            model.23.dfl            Identity     False           0                  []         -         -              -
  348    model.23.one2one_cv2.0.0.conv.weight              Conv2d     False      147456     [64, 256, 3, 3] -0.000363    0.0176        float32
  349      model.23.one2one_cv2.0.0.bn.weight         BatchNorm2d     False          64                [64]      1.63     0.851        float32
  349        model.23.one2one_cv2.0.0.bn.bias         BatchNorm2d     False          64                [64]   -0.0302      1.38        float32
  350            model.23.one2one_cv2.0.0.act                SiLU     False           0                  []         -         -              -
  351    model.23.one2one_cv2.0.1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00261     0.037        float32
  352      model.23.one2one_cv2.0.1.bn.weight         BatchNorm2d     False          64                [64]      1.27      1.11        float32
  352        model.23.one2one_cv2.0.1.bn.bias         BatchNorm2d     False          64                [64]    -0.386      1.13        float32
  353         model.23.one2one_cv2.0.2.weight              Conv2d     False         256       [4, 64, 1, 1]    0.0366      0.15        float32
  353           model.23.one2one_cv2.0.2.bias              Conv2d     False           4                 [4]     0.838    0.0901        float32
  354    model.23.one2one_cv2.1.0.conv.weight              Conv2d     False      294912     [64, 512, 3, 3] -0.000189   0.00984        float32
  355      model.23.one2one_cv2.1.0.bn.weight         BatchNorm2d     False          64                [64]      1.68     0.647        float32
  355        model.23.one2one_cv2.1.0.bn.bias         BatchNorm2d     False          64                [64]    -0.216     0.991        float32
  356    model.23.one2one_cv2.1.1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00152    0.0288        float32
  357      model.23.one2one_cv2.1.1.bn.weight         BatchNorm2d     False          64                [64]      1.01     0.872        float32
  357        model.23.one2one_cv2.1.1.bn.bias         BatchNorm2d     False          64                [64]     -0.16     0.856        float32
  358         model.23.one2one_cv2.1.2.weight              Conv2d     False         256       [4, 64, 1, 1]    0.0524     0.152        float32
  358           model.23.one2one_cv2.1.2.bias              Conv2d     False           4                 [4]      1.21     0.142        float32
  359    model.23.one2one_cv2.2.0.conv.weight              Conv2d     False      294912     [64, 512, 3, 3] -0.000218   0.00868        float32
  360      model.23.one2one_cv2.2.0.bn.weight         BatchNorm2d     False          64                [64]      1.51      0.45        float32
  360        model.23.one2one_cv2.2.0.bn.bias         BatchNorm2d     False          64                [64]   -0.0657     0.611        float32
  361    model.23.one2one_cv2.2.1.conv.weight              Conv2d     False       36864      [64, 64, 3, 3]  -0.00165    0.0257        float32
  362      model.23.one2one_cv2.2.1.bn.weight         BatchNorm2d     False          64                [64]      1.14     0.716        float32
  362        model.23.one2one_cv2.2.1.bn.bias         BatchNorm2d     False          64                [64]   -0.0302     0.727        float32
  363         model.23.one2one_cv2.2.2.weight              Conv2d     False         256       [4, 64, 1, 1]     0.044     0.142        float32
  363           model.23.one2one_cv2.2.2.bias              Conv2d     False           4                 [4]      1.51     0.238        float32
  364  model.23.one2one_cv3.0.0.0.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]   0.00968    0.0565        float32
  365    model.23.one2one_cv3.0.0.0.bn.weight         BatchNorm2d     False         256               [256]     0.491     0.256        float32
  365      model.23.one2one_cv3.0.0.0.bn.bias         BatchNorm2d     False         256               [256]      1.08     0.726        float32
  366          model.23.one2one_cv3.0.0.0.act                SiLU     False           0                  []         -         -              -
  367  model.23.one2one_cv3.0.0.1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00156    0.0437        float32
  368    model.23.one2one_cv3.0.0.1.bn.weight         BatchNorm2d     False         256               [256]     0.662      0.29        float32
  368      model.23.one2one_cv3.0.0.1.bn.bias         BatchNorm2d     False         256               [256]     0.069     0.601        float32
  369  model.23.one2one_cv3.0.1.0.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]   0.00385    0.0613        float32
  370    model.23.one2one_cv3.0.1.0.bn.weight         BatchNorm2d     False         256               [256]     0.806     0.552        float32
  370      model.23.one2one_cv3.0.1.0.bn.bias         BatchNorm2d     False         256               [256]     0.655       0.8        float32
  371  model.23.one2one_cv3.0.1.1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1] -0.000823    0.0451        float32
  372    model.23.one2one_cv3.0.1.1.bn.weight         BatchNorm2d     False         256               [256]      1.93      1.06        float32
  372      model.23.one2one_cv3.0.1.1.bn.bias         BatchNorm2d     False         256               [256]     0.683     0.911        float32
  373         model.23.one2one_cv3.0.2.weight              Conv2d     False       20480     [80, 256, 1, 1]  -0.00801    0.0555        float32
  373           model.23.one2one_cv3.0.2.bias              Conv2d     False          80                [80]     -11.7     0.264        float32
  374  model.23.one2one_cv3.1.0.0.conv.weight              Conv2d     False        4608      [512, 1, 3, 3]    0.0058    0.0405        float32
  375    model.23.one2one_cv3.1.0.0.bn.weight         BatchNorm2d     False         512               [512]     0.789     0.304        float32
  375      model.23.one2one_cv3.1.0.0.bn.bias         BatchNorm2d     False         512               [512]     0.356     0.408        float32
  376  model.23.one2one_cv3.1.0.1.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  -0.00167     0.036        float32
  377    model.23.one2one_cv3.1.0.1.bn.weight         BatchNorm2d     False         256               [256]     0.857     0.212        float32
  377      model.23.one2one_cv3.1.0.1.bn.bias         BatchNorm2d     False         256               [256]    0.0638     0.546        float32
  378  model.23.one2one_cv3.1.1.0.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]   0.00131    0.0556        float32
  379    model.23.one2one_cv3.1.1.0.bn.weight         BatchNorm2d     False         256               [256]     0.878     0.628        float32
  379      model.23.one2one_cv3.1.1.0.bn.bias         BatchNorm2d     False         256               [256]     0.285     0.558        float32
  380  model.23.one2one_cv3.1.1.1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]   -0.0011    0.0413        float32
  381    model.23.one2one_cv3.1.1.1.bn.weight         BatchNorm2d     False         256               [256]      1.84      1.13        float32
  381      model.23.one2one_cv3.1.1.1.bn.bias         BatchNorm2d     False         256               [256]      0.75     0.784        float32
  382         model.23.one2one_cv3.1.2.weight              Conv2d     False       20480     [80, 256, 1, 1]   -0.0105    0.0574        float32
  382           model.23.one2one_cv3.1.2.bias              Conv2d     False          80                [80]     -10.4      0.21        float32
  383  model.23.one2one_cv3.2.0.0.conv.weight              Conv2d     False        4608      [512, 1, 3, 3]   0.00326     0.036        float32
  384    model.23.one2one_cv3.2.0.0.bn.weight         BatchNorm2d     False         512               [512]     0.955      0.24        float32
  384      model.23.one2one_cv3.2.0.0.bn.bias         BatchNorm2d     False         512               [512]    0.0792     0.176        float32
  385  model.23.one2one_cv3.2.0.1.conv.weight              Conv2d     False      131072    [256, 512, 1, 1]  -0.00157    0.0396        float32
  386    model.23.one2one_cv3.2.0.1.bn.weight         BatchNorm2d     False         256               [256]     0.935     0.124        float32
  386      model.23.one2one_cv3.2.0.1.bn.bias         BatchNorm2d     False         256               [256]    0.0771     0.375        float32
  387  model.23.one2one_cv3.2.1.0.conv.weight              Conv2d     False        2304      [256, 1, 3, 3]  -0.00485    0.0465        float32
  388    model.23.one2one_cv3.2.1.0.bn.weight         BatchNorm2d     False         256               [256]     0.872     0.516        float32
  388      model.23.one2one_cv3.2.1.0.bn.bias         BatchNorm2d     False         256               [256]     0.144      0.31        float32
  389  model.23.one2one_cv3.2.1.1.conv.weight              Conv2d     False       65536    [256, 256, 1, 1]  -0.00086    0.0415        float32
  390    model.23.one2one_cv3.2.1.1.bn.weight         BatchNorm2d     False         256               [256]      1.49     0.714        float32
  390      model.23.one2one_cv3.2.1.1.bn.bias         BatchNorm2d     False         256               [256]     0.722     0.771        float32
  391         model.23.one2one_cv3.2.2.weight              Conv2d     False       20480     [80, 256, 1, 1]   -0.0114    0.0624        float32
  391           model.23.one2one_cv3.2.2.bias              Conv2d     False          80                [80]     -9.01     0.142        float32
YOLO26l summary: 392 layers, 26,299,704 parameters, 0 gradients, 93.8 GFLOPs

(392, 26299704, 0, 93.7977856)

Section 4: Training Configuration and Execution

Purpose: This is the computational core of the notebook. It defines the hyperparameters that control how the model learns and executes the training loop.

Key Activities:

• Hyperparameter Definition: Sets critical parameters such as epochs (100), batch_size (96), and img_size (640).
• Augmentation Strategy: Configures geometric and color-space augmentations (Mosaic, HSV shifts, Scale) to artificially increase dataset diversity and prevent overfitting.
• Smart Resume Logic: The script checks if a training run was interrupted. If valid weights exist in Google Drive, it automatically resumes training from the last checkpoint; otherwise, it starts a new session.
• Training Loop: Calls model.train() to begin the backpropagation process, optimizing the weights to detect the five life stages.

# A unique identifier for this specific training run, used for naming output folders.
experiment_name = 'run_v1'
# The total number of times the model will iterate over the entire training dataset.
epochs = 100
# The number of images processed in a single forward/backward pass of the model.
batch_size = 24
# The resolution (in pixels) to which all input images will be resized before training.
img_size = 640
# The number of consecutive epochs with no improvement in validation metrics before training is stopped early.
patience = 15

# The initial step size for the optimizer's weight updates.
learning_rate = 0.01
# The optimization algorithm to be used. 'auto' allows Ultralytics to select a suitable default.
optimizer_type = 'MuSGD'

# Probability of applying the mosaic augmentation, which combines four images into one.
mosaic_probability = 1.0
# The degree of random hue shift applied in HSV color-space augmentation.
hsv_hue_fraction = 0.015
# The degree of random saturation shift applied in HSV color-space augmentation.
hsv_saturation_fraction = 0.7
# The degree of random brightness shift applied in HSV color-space augmentation.
hsv_brightness_fraction = 0.0
# The range (in degrees) of random rotation applied to images.
rotation_degrees = 0.0
# The probability of vertically flipping an image.
flip_vertical_probability = 0.0
# The probability of horizontally flipping an image.
flip_horizontal_probability = 0.0
# The gain for applying random scaling (zoom in/out) to images.
scale_gain = 0.5

# A boolean flag to force the training to start from scratch, ignoring any existing checkpoints.
force_restart = False
# The default path to the 'last.pt' checkpoint file for automatic training resumption.
auto_resume_path = os.path.join(weights_path, 'last.pt')
# An optional path to a specific weight file to resume from, overriding the auto-resume logic if set.
specific_weight_path = ""

# Prints a formatted summary of the key training and augmentation configurations for user verification.
print("-" * 40)
print("Training Configuration")
print(f"{'Experiment Name':<25} : {experiment_name}")
print(f"{'Epochs':<25} : {epochs}")
print(f"{'Batch Size':<25} : {batch_size}")
print(f"{'Image Size':<25} : {img_size}")
print(f"{'Learning Rate':<25} : {learning_rate}")
print("-" * 40)
print("Augmentation Strategy")
print(f"{'Mosaic Probability':<25} : {mosaic_probability}")
print(f"{'Hue Fraction':<25} : {hsv_hue_fraction}")
print(f"{'Saturation Fraction':<25} : {hsv_saturation_fraction}")
print(f"{'Brightness Fraction':<25} : {hsv_brightness_fraction}")
print(f"{'Rotation Degrees':<25} : {rotation_degrees}")
print(f"{'Vertical Flip Prob':<25} : {flip_vertical_probability}")
print(f"{'Horizontal Flip Prob':<25} : {flip_horizontal_probability}")
print("-" * 40)
print(f"{'Auto Resume Path':<25} : {auto_resume_path}")

----------------------------------------
Training Configuration
Experiment Name           : run_v1
Epochs                    : 100
Batch Size                : 24
Image Size                : 640
Learning Rate             : 0.01
----------------------------------------
Augmentation Strategy
Mosaic Probability        : 1.0
Hue Fraction              : 0.015
Saturation Fraction       : 0.7
Brightness Fraction       : 0.0
Rotation Degrees          : 0.0
Vertical Flip Prob        : 0.0
Horizontal Flip Prob      : 0.0
----------------------------------------
Auto Resume Path          : /content/drive/MyDrive/Mortiscope Models/YOLO26/weights/last.pt

import time

# Checkpoint Resumption Logic

# Initializes the default training mode to start a new session.
resume_mode = False
# Sets the default weights to the pre-trained model file.
weights_to_load = model_filename

# Checks for a specific, manually provided weight file to resume from.
if specific_weight_path and os.path.exists(specific_weight_path):
    print(f"Manual override detected.\nResuming from specific weight: {specific_weight_path}")
    resume_mode = True
    weights_to_load = specific_weight_path
# If no manual override is given, checks for the default auto-resume checkpoint file.
elif os.path.exists(auto_resume_path):
    # Handles the case where a checkpoint exists but a fresh start is explicitly required.
    if force_restart:
        print(f"Previous run found at {auto_resume_path}, but 'force_restart' is True.")
        print("Starting new training...")
        resume_mode = False
    # If a checkpoint exists and a fresh start is not forced, sets up auto-resumption.
    else:
        print(f"Previous run detected.\nAuto-resuming from: {auto_resume_path}")
        resume_mode = True
        weights_to_load = auto_resume_path
# If no checkpoints are found, configures the script to start a new training session.
else:
    print("No previous run found. Starting new training...")


# Training Execution
print("\nInitializing training loop...")
# Records the timestamp at the beginning of the training process to measure total duration.
start_time = time.time()

# Executes the appropriate training command based on whether a session is being resumed.
if resume_mode:
    # When resuming, the `resume` argument is used, which automatically loads the model's state, optimizer state, and last epoch from the checkpoint.
    results = model.train(
        resume=weights_to_load,
        project=project_path,
        name=experiment_name
    )
else:
    # For a new run, all hyperparameters and augmentation settings are passed explicitly.
    results = model.train(
        # Core Parameters
        data=yaml_path,
        epochs=epochs,
        imgsz=img_size,
        batch=batch_size,
        device=0,
        patience=patience,
        # Project and Checkpointing
        save=True,
        save_period=1,
        project=project_path,
        name=experiment_name,
        exist_ok=True,
        plots=True,
        # Optimizer Settings
        lr0=learning_rate,
        optimizer=optimizer_type,
        # Augmentation Parameters
        mosaic=mosaic_probability,
        hsv_h=hsv_hue_fraction,
        hsv_s=hsv_saturation_fraction,
        hsv_v=hsv_brightness_fraction,
        degrees=rotation_degrees,
        flipud=flip_vertical_probability,
        fliplr=flip_horizontal_probability,
        scale=scale_gain
    )

# Records the timestamp at the end of the training process.
end_time = time.time()

# Calculates the total training duration and formats it into hours, minutes, and seconds.
duration_seconds = end_time - start_time
hours = int(duration_seconds // 3600)
minutes = int((duration_seconds % 3600) // 60)
seconds = int(duration_seconds % 60)

# Prints a final summary of the completed training session and its duration.
print("\n" + "-"*30)
print("Training complete.")
print(f"Total Time: {hours}h {minutes}m {seconds}s")
print("-"*30)

No previous run found. Starting new training...

Initializing training loop...
Ultralytics 8.4.5 🚀 Python-3.12.12 torch-2.9.0+cu126 CUDA:0 (NVIDIA L4, 22693MiB)
engine/trainer: agnostic_nms=False, amp=True, angle=1.0, augment=False, auto_augment=randaugment, batch=24, bgr=0.0, box=7.5, cache=False, cfg=None, classes=None, close_mosaic=10, cls=0.5, compile=False, conf=None, copy_paste=0.0, copy_paste_mode=flip, cos_lr=False, cutmix=0.0, data=/content/data.yaml, degrees=0.0, deterministic=True, device=0, dfl=1.5, dnn=False, dropout=0.0, dynamic=False, embed=None, epochs=100, erasing=0.4, exist_ok=True, fliplr=0.0, flipud=0.0, format=torchscript, fraction=1.0, freeze=None, half=False, hsv_h=0.015, hsv_s=0.7, hsv_v=0.0, imgsz=640, int8=False, iou=0.7, keras=False, kobj=1.0, line_width=None, lr0=0.01, lrf=0.01, mask_ratio=4, max_det=300, mixup=0.0, mode=train, model=yolo26l.pt, momentum=0.937, mosaic=1.0, multi_scale=0.0, name=run_v1, nbs=64, nms=False, opset=None, optimize=False, optimizer=MuSGD, overlap_mask=True, patience=15, perspective=0.0, plots=True, pose=12.0, pretrained=True, profile=False, project=/content/drive/MyDrive/Mortiscope Models/YOLO26, rect=False, resume=False, retina_masks=False, rle=1.0, save=True, save_conf=False, save_crop=False, save_dir=/content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1, save_frames=False, save_json=False, save_period=1, save_txt=False, scale=0.5, seed=0, shear=0.0, show=False, show_boxes=True, show_conf=True, show_labels=True, simplify=True, single_cls=False, source=None, split=val, stream_buffer=False, task=detect, time=None, tracker=botsort.yaml, translate=0.1, val=True, verbose=True, vid_stride=1, visualize=False, warmup_bias_lr=0.1, warmup_epochs=3.0, warmup_momentum=0.8, weight_decay=0.0005, workers=8, workspace=None
Overriding model.yaml nc=80 with nc=5

                   from  n    params  module                                       arguments                     
  0                  -1  1      1856  ultralytics.nn.modules.conv.Conv             [3, 64, 3, 2]                 
  1                  -1  1     73984  ultralytics.nn.modules.conv.Conv             [64, 128, 3, 2]               
  2                  -1  2    173824  ultralytics.nn.modules.block.C3k2            [128, 256, 2, True, 0.25]     
  3                  -1  1    590336  ultralytics.nn.modules.conv.Conv             [256, 256, 3, 2]              
  4                  -1  2    691712  ultralytics.nn.modules.block.C3k2            [256, 512, 2, True, 0.25]     
  5                  -1  1   2360320  ultralytics.nn.modules.conv.Conv             [512, 512, 3, 2]              
  6                  -1  2   2234368  ultralytics.nn.modules.block.C3k2            [512, 512, 2, True]           
  7                  -1  1   2360320  ultralytics.nn.modules.conv.Conv             [512, 512, 3, 2]              
  8                  -1  2   2234368  ultralytics.nn.modules.block.C3k2            [512, 512, 2, True]           
  9                  -1  1    656896  ultralytics.nn.modules.block.SPPF            [512, 512, 5, 3, True]        
 10                  -1  2   1455616  ultralytics.nn.modules.block.C2PSA           [512, 512, 2]                 
 11                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          
 12             [-1, 6]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 13                  -1  2   2496512  ultralytics.nn.modules.block.C3k2            [1024, 512, 2, True]          
 14                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          
 15             [-1, 4]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 16                  -1  2    756736  ultralytics.nn.modules.block.C3k2            [1024, 256, 2, True]          
 17                  -1  1    590336  ultralytics.nn.modules.conv.Conv             [256, 256, 3, 2]              
 18            [-1, 13]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 19                  -1  2   2365440  ultralytics.nn.modules.block.C3k2            [768, 512, 2, True]           
 20                  -1  1   2360320  ultralytics.nn.modules.conv.Conv             [512, 512, 3, 2]              
 21            [-1, 10]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 22                  -1  1   1974784  ultralytics.nn.modules.block.C3k2            [1024, 512, 1, True, 0.5, True]
 23        [16, 19, 22]  1   2806326  ultralytics.nn.modules.head.Detect           [5, 1, True, [256, 512, 512]] 

YOLO26l summary: 392 layers, 26,184,054 parameters, 26,184,054 gradients, 93.2 GFLOPs
Transferred 1080/1092 items from pretrained weights

AMP: running Automatic Mixed Precision (AMP) checks...
AMP: checks passed ✅
train: Fast image access ✅ (ping: 0.0±0.0 ms, read: 1634.4±574.3 MB/s, size: 59.0 KB)
train: Scanning /content/Complementary-1/train/labels.cache... 14102 images, 347 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 14102/14102 2.3Git/s 0.0s
albumentations: Blur(p=0.01, blur_limit=(3, 7)), MedianBlur(p=0.01, blur_limit=(3, 7)), ToGray(p=0.01, method='weighted_average', num_output_channels=3), CLAHE(p=0.01, clip_limit=(1.0, 4.0), tile_grid_size=(8, 8))
val: Fast image access ✅ (ping: 0.0±0.0 ms, read: 1107.9±883.0 MB/s, size: 44.3 KB)
val: Scanning /content/Complementary-1/valid/labels.cache... 896 images, 36 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 896/896 41.3Mit/s 0.0s
Plotting labels to /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/labels.jpg... 
optimizer: MuSGD(lr=0.01, momentum=0.937) with parameter groups 0 weight(decay=0.0), 0 weight(decay=0.0005625000000000001), 24 bias(decay=0.0)

Image sizes 640 train, 640 val
Using 8 dataloader workers
Logging results to /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1
Starting training for 100 epochs...

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      1/100      17.6G      1.756      3.078   0.008582         62        640: 100% ━━━━━━━━━━━━ 588/588 1.2s/it 11:24
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.1it/s 17.0s
                   all        896       2844      0.708      0.475      0.541      0.361

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      2/100      17.5G      1.653      1.588   0.008149         55        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 9:19
   History Saved: 002_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.3s
                   all        896       2844       0.66      0.528      0.585      0.377

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      3/100      17.2G      1.733       1.69   0.008931         51        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 9:06
   History Saved: 003_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.3s
                   all        896       2844      0.626      0.546      0.605      0.355

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      4/100      17.8G      1.768      1.721   0.009022        212        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 004_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.676      0.537      0.591      0.369

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      5/100      17.5G      1.732      1.612   0.008864         26        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 005_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.704      0.566      0.619       0.39

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      6/100      18.4G      1.698      1.521   0.008581         59        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 006_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.766      0.505      0.613      0.389

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      7/100      18.1G       1.66      1.457     0.0084         38        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 007_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844        0.7      0.591      0.647      0.426

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      8/100      16.7G      1.647      1.412   0.008352         47        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:37
   History Saved: 008_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.652      0.567      0.624      0.399

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
      9/100      18.5G      1.633      1.357   0.008182         64        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 009_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.783      0.483      0.598      0.406

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     10/100      17.3G      1.615      1.316   0.007985         41        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 010_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844       0.69      0.582      0.644      0.428

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     11/100      17.8G      1.606      1.279   0.007941         29        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 011_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.714      0.575      0.656       0.43

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     12/100      18.4G      1.583      1.235   0.007837         62        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 012_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844       0.67      0.603      0.656      0.443

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     13/100      17.4G      1.568      1.217   0.007715         27        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:37
   History Saved: 013_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.658      0.609      0.668      0.452

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     14/100      17.2G      1.558      1.193   0.007633         69        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 014_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.686      0.592      0.658      0.444

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     15/100      16.6G       1.54       1.15   0.007519         32        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 015_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.3s
                   all        896       2844      0.749      0.625      0.696      0.469

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     16/100      18.4G      1.548       1.14   0.007483         63        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 016_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.717      0.624       0.69      0.469

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     17/100      18.1G      1.533      1.119   0.007428         36        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 017_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.697       0.53      0.616      0.421

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     18/100      17.4G      1.514      1.101   0.007313         19        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:39
   History Saved: 018_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.672      0.609      0.648      0.438

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     19/100      17.5G      1.504      1.063   0.007284         26        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 019_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.688      0.563      0.647      0.439

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     20/100      17.5G      1.486      1.035    0.00717         26        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 020_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.692      0.611      0.662      0.458

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     21/100      18.2G      1.472      1.001   0.007081         36        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 021_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.713      0.619      0.683      0.472

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     22/100      17.3G      1.467          1   0.006978         44        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 022_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.731       0.61      0.687      0.471

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     23/100      17.7G      1.456     0.9634   0.006957         55        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 023_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844        0.7       0.66      0.694      0.473

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     24/100      16.6G       1.45     0.9486   0.006922         45        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:37
   History Saved: 024_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.685      0.633      0.688      0.473

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     25/100      17.2G      1.437     0.9489   0.006776        121        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 025_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.719      0.645      0.693      0.475

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     26/100      17.3G      1.431     0.9191   0.006854         65        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 026_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.724      0.625      0.694      0.475

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     27/100      17.9G      1.417     0.8933   0.006743         46        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 027_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.727      0.619      0.691      0.478

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     28/100      17.6G      1.411     0.8865   0.006591         69        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 028_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.711      0.621      0.687       0.48

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     29/100      17.7G      1.408     0.8736   0.006599         44        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:39
   History Saved: 029_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.659      0.631      0.663      0.465

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     30/100      16.6G      1.389     0.8486   0.006506         42        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 030_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.712      0.645      0.697      0.486

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     31/100      18.1G       1.38     0.8378   0.006486         31        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:39
   History Saved: 031_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.702      0.619      0.675       0.47

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     32/100      17.6G      1.371      0.824   0.006509         25        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 032_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.716      0.614      0.678      0.474

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     33/100      17.1G       1.36     0.8094   0.006336         52        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:39
   History Saved: 033_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.707      0.627      0.677      0.474

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     34/100      17.7G      1.353     0.7981   0.006367         54        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 034_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.705      0.626       0.67      0.472

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     35/100      17.6G       1.35     0.7866   0.006313         41        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:39
   History Saved: 035_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.698      0.667      0.697      0.488

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     36/100      16.8G      1.333     0.7753   0.006232        113        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 036_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.666      0.665      0.676      0.474

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     37/100      16.8G      1.306      0.762   0.006032         37        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 037_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844       0.69       0.65      0.684      0.481

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     38/100      17.1G       1.31     0.7485   0.006164         43        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:37
   History Saved: 038_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.681      0.648      0.672      0.472

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     39/100      17.7G      1.293     0.7361   0.006087         21        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 039_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.728      0.619      0.689      0.485

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     40/100      17.3G      1.283     0.7183   0.005949         19        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 040_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.725       0.64      0.687      0.487

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     41/100      17.5G      1.276     0.7061   0.005795         91        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 041_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.716      0.641       0.68      0.481

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     42/100      17.4G      1.268     0.7037   0.005825         58        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 042_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.735      0.633      0.685      0.486

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     43/100      18.3G      1.253       0.69   0.005776        115        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:39
   History Saved: 043_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.732      0.626      0.684      0.484

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     44/100      18.2G      1.258     0.6893   0.005849         55        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 044_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.711      0.632      0.675      0.478

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     45/100      17.5G      1.241     0.6704   0.005634         44        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:39
   History Saved: 045_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.724      0.623      0.676      0.482

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     46/100      17.4G       1.23     0.6632    0.00569         34        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 046_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.719      0.638      0.686      0.486

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     47/100      18.2G      1.226     0.6551    0.00565         48        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 047_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.732      0.628      0.683      0.486

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     48/100      17.3G      1.218     0.6379   0.005575         42        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:39
   History Saved: 048_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.705      0.642      0.671      0.475

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     49/100      16.6G      1.199     0.6305   0.005587         77        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 049_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.1s
                   all        896       2844      0.723      0.636      0.677       0.48

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size
     50/100      17.4G      1.198     0.6258   0.005519         73        640: 100% ━━━━━━━━━━━━ 588/588 1.1it/s 8:38
   History Saved: 050_epoch.pt
                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.7it/s 11.2s
                   all        896       2844      0.752      0.618      0.679      0.484

EarlyStopping: Training stopped early as no improvement observed in last 15 epochs. Best results observed at epoch 35, best model saved as best.pt.

To update EarlyStopping(patience=15) pass a new patience value, i.e. `patience=300` or use `patience=0` to disable EarlyStopping.

50 epochs completed in 7.469 hours.
Optimizer stripped from /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/last.pt, 53.0MB
Optimizer stripped from /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.pt, 53.0MB

Validating /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.pt...
Ultralytics 8.4.5 🚀 Python-3.12.12 torch-2.9.0+cu126 CUDA:0 (NVIDIA L4, 22693MiB)
YOLO26l summary (fused): 190 layers, 24,749,595 parameters, 0 gradients, 86.1 GFLOPs

                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 19/19 1.9it/s 10.2s
                   all        896       2844      0.698      0.667      0.697      0.488
                 adult        176        593      0.742      0.548      0.594      0.327
              instar_1        222       1164      0.656      0.512      0.598      0.332
              instar_2        168        521      0.402       0.57      0.496      0.321
              instar_3        240        334      0.698      0.776      0.836      0.656
                  pupa        205        232      0.991      0.928      0.961      0.803

Speed: 0.2ms preprocess, 8.6ms inference, 0.0ms loss, 0.3ms postprocess per image
Results saved to /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1

------------------------------
Training complete.
Total Time: 7h 28m 36s
------------------------------

Section 5: Visual Performance Evaluation

Purpose: To assess the model's accuracy and reliability using visual metrics. Raw numbers are often insufficient; visualization helps identify specific classes or scenarios where the model struggles.

Key Activities:

• Training Metrics Plotting: Parses the results.csv log file to generate line charts for Box Loss, Classification Loss, Precision, Recall, and mAP (Mean Average Precision). This helps verify that the model is converging and not overfitting.
• Confusion Matrix Generation: Runs a validation pass on the best-performing model to generate a normalized Confusion Matrix. This heatmap visualizes how often the model confuses one life stage with another, providing insight into biological similarities affecting the AI.

import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns

# Constructs the full path to the results log file.
results_csv_path = os.path.join(project_path, experiment_name, 'results.csv')

# The number of data points to average over when applying a rolling mean.
smoothing_window_size = 5

# Sets a professional and consistent visual theme for all generated plots.
sns.set_theme(style="whitegrid", context="notebook", font_scale=1.1)
# Increases the resolution of the output figures for better clarity.
plt.rcParams['figure.dpi'] = 120

# Defines a perceptually uniform colormap to derive a consistent color palette.
color_map = plt.get_cmap('plasma')

# Assigns specific colors from the colormap to different metrics for consistency.
color_train = color_map(0.0)       # Color for training metrics.
color_val   = color_map(0.6)       # Color for validation metrics.
color_prec  = color_map(0.25)      # Color for the precision curve.
color_rec   = color_map(0.75)      # Color for the recall curve.
color_map_metric = color_map(0.5)  # Color for the mAP metric.
color_lr    = color_map(0.05)      # Color for the learning rate schedule.


def plot_training_results(csv_file_path):
    """
    Reads a results.csv file and generates a 2x3 grid of training performance plots.

    Args:
        csv_file_path (str): The full path to the results.csv file.
    """
    # Verifies the existence of the results file before attempting to read it.
    if not os.path.exists(csv_file_path):
        print(f"Error: Could not find results at {csv_file_path}")
        return

    # Reads the CSV data into a pandas DataFrame.
    df = pd.read_csv(csv_file_path)
    # Cleans up column names by removing leading/trailing whitespace.
    df.columns = df.columns.str.strip()

    # Creates a Matplotlib figure and an array of 2x3 subplots (axes).
    figure, axis_array = plt.subplots(2, 3, figsize=(18, 10))

    # Defines the mapping of DataFrame columns to their respective plot titles for the three primary loss functions.
    loss_map = [
        ('train/box_loss', 'val/box_loss', 'Box Loss'),
        ('train/cls_loss', 'val/cls_loss', 'Classification Loss'),
        ('train/dfl_loss', 'val/dfl_loss', 'Distribution Focal Loss')
    ]

    # Iterates through the loss map to generate the top row of plots.
    for i, (train_col, val_col, title) in enumerate(loss_map):
        axis = axis_array[0, i]

        # Plots the raw, noisy data with low opacity to serve as a background reference.
        sns.lineplot(data=df, x=df.index, y=train_col, ax=axis, color=color_train, alpha=0.15)
        sns.lineplot(data=df, x=df.index, y=val_col, ax=axis, color=color_val, alpha=0.15)

        # Overlays the smoothed data using a rolling mean for clearer trend visualization.
        sns.lineplot(x=df.index, y=df[train_col].rolling(smoothing_window_size).mean(),
                     ax=axis, color=color_train, linewidth=2.5, label='Train')
        sns.lineplot(x=df.index, y=df[val_col].rolling(smoothing_window_size).mean(),
                     ax=axis, color=color_val, linewidth=2.5, label='Validation')

        # Configures the title, labels, and legend for each loss plot.
        axis.set_title(title, color='#333333')
        axis.set_xlabel('Epochs')
        axis.set_ylabel('Loss Value')
        axis.legend()

    # Precision and Recall Plot
    axis_precision_recall = axis_array[1, 0]
    sns.lineplot(x=df.index, y=df['metrics/precision(B)'].rolling(smoothing_window_size).mean(),
                 ax=axis_precision_recall, color=color_prec, label='Precision')
    sns.lineplot(x=df.index, y=df['metrics/recall(B)'].rolling(smoothing_window_size).mean(),
                 ax=axis_precision_recall, color=color_rec, label='Recall')

    axis_precision_recall.set_title('Precision & Recall')
    axis_precision_recall.set_xlabel('Epochs')
    axis_precision_recall.set_ylabel('Score')
    axis_precision_recall.set_ylim(0, 1)

    # Mean Average Precision (mAP) Plot
    axis_map = axis_array[1, 1]
    sns.lineplot(x=df.index, y=df['metrics/mAP50(B)'].rolling(smoothing_window_size).mean(),
                 ax=axis_map, color=color_map_metric, linewidth=2.5, label='mAP @ 0.50')

    axis_map.set_title('Mean Average Precision (IoU=0.50)')
    axis_map.set_xlabel('Epochs')
    axis_map.set_ylabel('Score')
    axis_map.set_ylim(0, 1)
    # Fills the area under the mAP curve to visually emphasize the performance metric.
    axis_map.fill_between(df.index, df['metrics/mAP50(B)'].rolling(smoothing_window_size).mean(),
                          color=color_map_metric, alpha=0.1)

    # Learning Rate Schedule Plot
    axis_learning_rate = axis_array[1, 2]
    sns.lineplot(x=df.index, y=df['lr/pg0'], ax=axis_learning_rate, color=color_lr, linestyle='--')

    axis_learning_rate.set_title('Learning Rate Schedule')
    axis_learning_rate.set_xlabel('Epochs')
    axis_learning_rate.set_ylabel('Learning Rate')

    # Adjusts the spacing between subplots to prevent labels from overlapping.
    plt.tight_layout(
        pad=3.0,
        w_pad=4.0,
        h_pad=5.0
    )
    # Renders and displays the final, complete figure.
    plt.show()

# Executes the plotting function with the path to the results file.
plot_training_results(results_csv_path)

import numpy as np

# Defines the path to the best-performing model weights saved during training.
best_weight_path = os.path.join(project_path, experiment_name, 'weights', 'best.pt')
# Defines the output path for the final, publication-quality confusion matrix image.
output_image_path = os.path.join(project_path, experiment_name, 'confusion_matrix.png')

# Ensures that the script only runs if the best weight file exists.
if os.path.exists(best_weight_path):
    print(f"Loading weights from: {best_weight_path}")
    # Instantiates a new YOLO model object using the best saved weights.
    validation_model = YOLO(best_weight_path)

    # Runs a validation pass on the model.
    validation_metrics = validation_model.val(
        data=yaml_path,
        split='val',
        plots=True,
        device=0,
        batch=batch_size,
        conf=0.001
    )

    # Extracts the raw confusion matrix (a NumPy array) from the results.
    raw_matrix = validation_metrics.confusion_matrix.matrix
    num_classes = 5
    # Slices the matrix to ensure it only contains the defined classes.
    matrix_data = raw_matrix[:num_classes, :num_classes]

    # Retrieves the class names from the validation results and formats them.
    raw_names = list(validation_metrics.names.values())[:num_classes]
    class_names = [name.replace('_', ' ').title() for name in raw_names]

    # Normalizes the confusion matrix by rows to calculate recall scores.
    row_sums = matrix_data.sum(axis=1, keepdims=True)
    # Replaces zero sums with a small number to avoid division-by-zero errors for classes that may not have appeared in the validation set.
    row_sums[row_sums == 0] = 1e-9
    matrix_normalized = matrix_data / row_sums

    # Initializes a high-resolution figure for the plot.
    plt.figure(figsize=(16, 12), dpi=300)
    sns.set_theme(style="white", font_scale=1.1)

    # Creates the heatmap using Seaborn, configuring annotations, colormap, and labels.
    axis = sns.heatmap(
        matrix_normalized,
        annot=True,                 # Displays the numerical value in each cell.
        annot_kws={"size": 14},     # Sets the font size for annotations.
        fmt='.2f',                  # Formats annotations to two decimal places.
        cmap='Blues',               # Sets the color scheme.
        xticklabels=class_names,
        yticklabels=class_names,
        vmin=0.0, vmax=1.0,         # Fixes the color bar range from 0 to 1.
        square=True,                # Enforces square cells for better proportionality.
        linewidths=2.5,
        linecolor='white',
        cbar_kws={
            'shrink': 0.6,          # Adjusts the size of the color bar.
            'pad': 0.04
        }
    )

    # Configures the color bar label to clarify that the values represent recall.
    cbar = axis.collections[0].colorbar
    cbar.set_label('Recall (Sensitivity)', labelpad=30, fontsize=14)

    # Sets the main title and axis labels with appropriate padding.
    plt.title('Confusion Matrix', fontsize=20, pad=30)
    plt.xlabel('Predicted Class', fontsize=16, labelpad=25)
    plt.ylabel('Actual Class', fontsize=16, labelpad=25)

    # Adjusts tick label appearance for clarity.
    plt.xticks(rotation=0, fontsize=13)
    plt.yticks(rotation=0, fontsize=13)

    # Adjusts subplot parameters to give a tight layout.
    plt.tight_layout(pad=5.0)

    # Saves the final figure to the specified output path.
    plt.savefig(output_image_path, dpi=300, bbox_inches='tight')
    print(f"Confusion Matrix saved to: {output_image_path}")

    # Displays the plot in the notebook output.
    plt.show()

else:
    # Prints an error message if the required 'best.pt' file is not found.
    print(f"Error: Best weights not found at {best_weight_path}")

Loading weights from: /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.pt

Ultralytics 8.4.5 🚀 Python-3.12.12 torch-2.9.0+cu126 CUDA:0 (NVIDIA L4, 22693MiB)
YOLO26l summary (fused): 190 layers, 24,749,595 parameters, 0 gradients, 86.1 GFLOPs
val: Fast image access ✅ (ping: 0.0±0.0 ms, read: 1336.2±751.5 MB/s, size: 42.6 KB)
val: Scanning /content/Complementary-1/valid/labels.cache... 896 images, 36 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 896/896 375.8Mit/s 0.0s

                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 38/38 2.3it/s 16.4s
                   all        896       2844      0.698      0.666      0.697      0.488
                 adult        176        593      0.744      0.548      0.594      0.328
              instar_1        222       1164      0.656       0.51      0.598      0.332
              instar_2        168        521      0.404       0.57      0.496      0.321
              instar_3        240        334      0.698      0.776      0.836      0.655
                  pupa        205        232      0.991      0.927      0.961      0.804

Speed: 1.0ms preprocess, 14.8ms inference, 0.0ms loss, 0.1ms postprocess per image
Results saved to /content/runs/detect/val
Confusion Matrix saved to: /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/confusion_matrix.png

Section 6: Benchmarking and SAHI Integration

Purpose: To evaluate the model’s operational efficiency (speed) and prepare it for advanced inference scenarios involving high-resolution imagery.

Key Activities:

• Speed Benchmarking: Runs a validation pass specifically to measure pre-processing, inference, and post-processing times. This calculates the estimated FPS (Frames Per Second) to determine if the model is suitable for real-time applications.
• SAHI Wrapper Initialization: Initializes the Sliced Aided Hyper Inference wrapper. Standard YOLO resizing can make small insects vanish in 4K images. SAHI solves this by slicing the image into smaller overlapping windows, performing inference on each, and stitching the results back together.

# Defines the path to the best-performing model weights from the training run.
best_weight_path = os.path.join(project_path, experiment_name, 'weights', 'best.pt')

# Ensures the script proceeds only if the required model weight file is found.
if os.path.exists(best_weight_path):
    print(f"Loading best model for benchmarking: {best_weight_path}")
    # Loads the best model weights into a YOLO object for evaluation.
    benchmark_model = YOLO(best_weight_path)

    # Runs a validation pass on the specified dataset split.
    metrics = benchmark_model.val(data=yaml_path, split='val', plots=False, device=0)

    # Extracts the speed dictionary, which contains timing information for different stages of the inference pipeline.
    speed_metrics = metrics.speed

    # Displays a formatted summary of the average inference speed metrics.
    print("\n" + "-"*45)
    print("Inference Speed Benchmark (Average per Image)")
    print("-"*45)
    print(f"{'Pre-process':<25} : {speed_metrics['preprocess']:.2f} ms")
    print(f"{'Inference (Model)':<25} : {speed_metrics['inference']:.2f} ms")
    print(f"{'Post-process (NMS)':<25} : {speed_metrics['postprocess']:.2f} ms")
    print("-" * 45)

    # Calculates the total latency by summing the timings of all pipeline stages.
    total_latency = sum(speed_metrics.values())
    print(f"{'Total Latency':<25} : {total_latency:.2f} ms")

    # Estimates the throughput in Frames Per Second (FPS) based on the total latency.
    fps = 1000 / total_latency
    print(f"{'Estimated FPS':<25} : {fps:.2f} fps")
    print("-"*45)

else:
    # Prints an error message if the model weights file could not be located.
    print("Error: Best weights file not found.")
    print("Please make sure that the training completed successfully.")

Loading best model for benchmarking: /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.pt

Ultralytics 8.4.5 🚀 Python-3.12.12 torch-2.9.0+cu126 CUDA:0 (NVIDIA L4, 22693MiB)
YOLO26l summary (fused): 190 layers, 24,749,595 parameters, 0 gradients, 86.1 GFLOPs
val: Fast image access ✅ (ping: 0.0±0.0 ms, read: 1523.4±525.3 MB/s, size: 38.5 KB)
val: Scanning /content/Complementary-1/valid/labels.cache... 896 images, 36 backgrounds, 0 corrupt: 100% ━━━━━━━━━━━━ 896/896 341.6Mit/s 0.0s

                 Class     Images  Instances      Box(P          R      mAP50  mAP50-95): 100% ━━━━━━━━━━━━ 56/56 3.9it/s 14.4s
                   all        896       2844      0.698      0.666      0.697      0.488
                 adult        176        593      0.743      0.548      0.594      0.328
              instar_1        222       1164      0.656      0.511      0.598      0.332
              instar_2        168        521      0.403       0.57      0.496      0.321
              instar_3        240        334      0.698      0.776      0.836      0.655
                  pupa        205        232      0.991      0.926      0.961      0.803

Speed: 0.6ms preprocess, 13.8ms inference, 0.0ms loss, 0.1ms postprocess per image

---------------------------------------------
Inference Speed Benchmark (Average per Image)
---------------------------------------------
Pre-process               : 0.65 ms
Inference (Model)         : 13.78 ms
Post-process (NMS)        : 0.11 ms
---------------------------------------------
Total Latency             : 14.54 ms
Estimated FPS             : 68.79 fps
---------------------------------------------

from sahi import AutoDetectionModel
from sahi.predict import get_sliced_prediction

# Defines the path to the best-performing model weights from the training run, which will be loaded into the SAHI wrapper.
best_weight_path = os.path.join(project_path, experiment_name, 'weights', 'best.pt')

# The height of each individual slice in pixels.
slice_height = 640
# The width of each individual slice in pixels.
slice_width = 640
# The percentage of overlap between adjacent slices vertically.
overlap_height_ratio = 0.2
# The percentage of overlap between adjacent slices horizontally.
overlap_width_ratio = 0.2

print(f"Initializing SAHI wrapper for: {best_weight_path}")

# Verifies that the model weight file exists before attempting to load it.
if os.path.exists(best_weight_path):
    # Initializes the SAHI AutoDetectionModel.
    detection_model = AutoDetectionModel.from_pretrained(
        model_type='ultralytics',     # Specifies the model architecture.
        model_path=best_weight_path,  # Provides the path to the custom-trained weights.
        confidence_threshold=0.25,    # Sets the minimum confidence for a detection to be considered valid.
        device="cuda:0"               # Assigns the model to a specific GPU device for inference.
    )

    # Prints a confirmation message summarizing the SAHI configuration.
    print("\n" + "-"*45)
    print("SAHI Model Ready")
    print("-" * 45)
    print(f"{'Slice Dimensions':<20} : {slice_height}x{slice_width}")
    print(f"{'Overlap Ratio':<20} : {overlap_height_ratio * 100}%")
    print(f"{'Confidence Thresh':<20} : 0.25")
    print("-" * 45)

else:
    # Handles the case where the required weight file is not found.
    print(f"Error: Weights not found at {best_weight_path}")
    print("Cannot initialize SAHI.")

Initializing SAHI wrapper for: /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.pt

---------------------------------------------
SAHI Model Ready
---------------------------------------------
Slice Dimensions     : 640x640
Overlap Ratio        : 20.0%
Confidence Thresh    : 0.25
---------------------------------------------

Section 7: Deployment Export and Interactive Demonstration

Purpose: To finalize the model for production deployment and provide a tangible demonstration of its capabilities on user-provided data.

Key Activities:

• ONNX Export: Converts the PyTorch model (.pt) to the ONNX (Open Neural Network Exchange) format. This format is hardware-agnostic and optimized for deployment on edge devices or web servers.
• Interactive Inference Pipeline: A comprehensive script that:
  1. Accepts a user-uploaded image.
  2. Detects the scene type (Macro vs. Field) to choose between Standard or Sliced inference.
  3. Filters outliers based on box area to reduce false positives.
  4. Draws bounding boxes and creates a summary legend of the detected entomological evidence.

# Defines the source path for the best PyTorch model weights and the desired target path and filename for the exported ONNX model.
source_weights = os.path.join(project_path, experiment_name, 'weights', 'best.pt')
target_filename = "yolo26_mortiscope.onnx"
target_path = os.path.join(project_path, experiment_name, 'weights', target_filename)

print(f"Loading weights from: {source_weights}")

# Verifies the existence of the source weight file before initiating the export process.
if os.path.exists(source_weights):
    # Loads the trained PyTorch model from the specified '.pt' file.
    model = YOLO(source_weights)

    # Executes the model export process with specific configurations.
    exported_path = model.export(
        format='onnx',      # Specifies the target export format as ONNX.
        dynamic=False,      # Exports the model with fixed input/output dimensions for performance.
        simplify=True,      # Applies the ONNX-Simplifier to optimize the model graph.
        opset=12            # Sets the ONNX operator set version for broad compatibility.
    )

    # After export, the file is renamed and moved to the final target location.
    if isinstance(exported_path, str):
        # The `export` method saves the file with a default name.
        shutil.move(exported_path, target_path)
        print("\n" + "-"*100)
        print(f"File Saved: {target_path}")
        print("-"*100)
    else:
        # Handles cases where the export process does not return a valid file path.
        print("Export returned unexpected format.")

else:
    # Provides an error message if the source PyTorch model weights are not found.
    print(f"Error: Could not find weights at {source_weights}")

Loading weights from: /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.pt

Ultralytics 8.4.5 🚀 Python-3.12.12 torch-2.9.0+cu126 CPU (Intel Xeon CPU @ 2.20GHz)
💡 ProTip: Export to OpenVINO format for best performance on Intel hardware. Learn more at https://docs.ultralytics.com/integrations/openvino/
YOLO26l summary (fused): 190 layers, 24,749,595 parameters, 0 gradients, 86.1 GFLOPs

PyTorch: starting from '/content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.pt' with input shape (1, 3, 640, 640) BCHW and output shape(s) (1, 300, 6) (50.5 MB)
ONNX: starting export with onnx 1.20.1 opset 12...

Exporting aten::index operator of advanced indexing in opset 12 is achieved by combination of multiple ONNX operators, including Reshape, Transpose, Concat, and Gather. If indices include negative values, the exported graph will produce incorrect results.
ONNX: slimming with onnxslim 0.1.82...
ONNX: export success ✅ 9.0s, saved as '/content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.onnx' (94.8 MB)
Export complete (10.2s)

Results saved to /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights
Predict:         yolo predict task=detect model=/content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.onnx imgsz=640 
Validate:        yolo val task=detect model=/content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/best.onnx imgsz=640 data=/content/data.yaml  
Visualize:       https://netron.app

----------------------------------------------------------------------------------------------------
File Saved: /content/drive/MyDrive/Mortiscope Models/YOLO26/run_v1/weights/yolo26_mortiscope.onnx
----------------------------------------------------------------------------------------------------

from collections import Counter, defaultdict

import cv2
import matplotlib.patches as mpatches
from google.colab import files
from sahi.prediction import ObjectPrediction

# Global Configuration

# The minimum confidence score required for a detection to be considered valid.
confidence_threshold = 0.25
# The target width in pixels for the final output visualization.
target_width = 3840
# The target height in pixels for the final output visualization.
target_height = 2160
# The resolution (Dots Per Inch) for the generated Matplotlib figure.
dpi = 100

# A standardized color palette for different insect life stages.
color_map = {
    "instar_1": "#eab308",
    "instar_2": "#84cc16",
    "instar_3": "#22c55e",
    "pupa":     "#f97316",
    "adult":    "#f43f5e"
}

# Defines the canonical order for presenting life stages in summaries and legends.
lifecycle_order = ["instar_1", "instar_2", "instar_3", "pupa", "adult"]


def hex_to_bgr(hex_color):
    """
    Converts a hexadecimal color string to a BGR tuple for use with OpenCV.

    Args:
        hex_color (str): The color in hexadecimal format (e.g., '#eab308').

    Returns:
        tuple: The color in BGR format (e.g., (8, 179, 234)).
    """
    hex_color = hex_color.lstrip('#')
    rgb = tuple(int(hex_color[i:i+2], 16) for i in (0, 2, 4))
    # OpenCV uses BGR order, so the RGB tuple is reversed.
    return (rgb[2], rgb[1], rgb[0])


def format_class_name(name):
    """
    Formats an internal class name into a human-readable, title-cased string.

    Args:
        name (str): The internal class name (e.g., 'instar_1').

    Returns:
        str: The formatted name (e.g., 'Instar 1').
    """
    return name.replace("_", " ").title()


def calculate_iou(box1, box2):
    """
    Calculates the Intersection over Union (IoU) of two bounding boxes.

    Args:
        box1 (sahi.prediction.BBox): The first bounding box.
        box2 (sahi.prediction.BBox): The second bounding box.

    Returns:
        float: The IoU score, a value between 0.0 and 1.0.
    """
    # Extracts coordinates for easier calculation.
    b1 = [box1.minx, box1.miny, box1.maxx, box1.maxy]
    b2 = [box2.minx, box2.miny, box2.maxx, box2.maxy]

    # Determines the coordinates of the intersection rectangle.
    x1 = max(b1[0], b2[0])
    y1 = max(b1[1], b2[1])
    x2 = min(b1[2], b2[2])
    y2 = min(b1[3], b2[3])

    # Computes the area of intersection.
    intersection = max(0, x2 - x1) * max(0, y2 - y1)
    # Computes the area of both bounding boxes.
    area1 = (b1[2] - b1[0]) * (b1[3] - b1[1])
    area2 = (b2[2] - b2[0]) * (b2[3] - b2[1])
    # Computes the area of the union.
    union = area1 + area2 - intersection

    if union == 0:
        return 0
    return intersection / union


def apply_class_agnostic_nms(predictions, iou_threshold=0.6):
    """
    Applies a custom class-agnostic Non-Maximum Suppression (NMS) to a list
    of object predictions to filter out highly overlapping boxes.

    Args:
        predictions (list[ObjectPrediction]): A list of SAHI ObjectPrediction objects.
        iou_threshold (float): The IoU threshold above which boxes are suppressed.

    Returns:
        list[ObjectPrediction]: A filtered list of object predictions.
    """
    # Sorts predictions by confidence score in descending order.
    sorted_preds = sorted(predictions, key=lambda x: x.score.value, reverse=True)
    kept_preds = []

    for current in sorted_preds:
        should_keep = True
        for kept in kept_preds:
            iou = calculate_iou(current.bbox, kept.bbox)
            if iou > iou_threshold:
                # Suppresses the current box if it has a high IoU with an already kept box.
                should_keep = False
                break
        if should_keep:
            kept_preds.append(current)

    return kept_preds


def detect_scene_type(sahi_model, image_path):
    """
    Analyzes an image to determine if it is a 'macro' (close-up) or 'field'
    (wide-angle) scene.

    This heuristic is based on the average relative area of objects detected
    in an initial, low-resolution pass. Large average areas suggest a macro shot.

    Args:
        sahi_model (sahi.AutoDetectionModel): The initialized SAHI model.
        image_path (str): The path to the image file.

    Returns:
        str: The detected scene type, either 'macro' or 'field'.
    """
    native_model = sahi_model.model
    results = native_model.predict(image_path, imgsz=640, conf=0.25, verbose=False)
    boxes = results[0].boxes

    if len(boxes) == 0:
        return "field"

    # Calculates the normalized area (width * height) of each detected box.
    areas = boxes.xywhn[:, 2] * boxes.xywhn[:, 3]
    avg_area = torch.mean(areas).item()

    # Classifies the scene based on a predefined area threshold.
    if avg_area > 0.015:
        return "macro"
    else:
        return "field"


def run_image_analysis():
    """
    Orchestrates the main image analysis workflow.

    This function handles the user file upload, selects an inference strategy
    based on the scene type, processes the detections, applies filtering, and
    generates a final visual report with annotations and a summary legend.
    """
    print("Click button to upload image:")
    uploaded_files = files.upload()

    if not uploaded_files:
        print("No file uploaded.")
        return

    for filename in uploaded_files.keys():
        print(f"\nProcessing {filename}")

        # Determines the appropriate inference strategy for the uploaded image.
        scene_type = detect_scene_type(detection_model, filename)

        image_cv = cv2.imread(filename)
        img_h, img_w, _ = image_cv.shape
        img_area = img_w * img_h

        object_prediction_list = []

        # Scene-Adaptive Inference
        if scene_type == "macro":
            # For close-up images, use the standard, non-sliced prediction method.
            native_model = detection_model.model

            results = native_model.predict(
                filename,
                conf=0.45,
                imgsz=640,
                augment=False,
                agnostic_nms=True, # Uses YOLO's built-in NMS.
                verbose=False
            )

            # Manually converts the native YOLO results into the SAHI ObjectPrediction format.
            for r in results:
                for box in r.boxes:
                    x1, y1, x2, y2 = box.xyxy[0].tolist()
                    score = box.conf[0].item()
                    cls_id = int(box.cls[0].item())
                    cls_name = native_model.names[cls_id]

                    obj = ObjectPrediction(
                        bbox=[x1, y1, x2, y2],
                        category_id=cls_id,
                        category_name=cls_name,
                        score=score
                    )
                    object_prediction_list.append(obj)

            use_outlier_filter = False

        else:
            # For wide-angle images, use SAHI's sliced prediction method.
            if img_w < 2500 or img_h < 2500:
                current_slice_size = 160
                current_overlap = 0.35
            else:
                current_slice_size = 320
                current_overlap = 0.25

            result = get_sliced_prediction(
                filename,
                detection_model,
                slice_height=current_slice_size,
                slice_width=current_slice_size,
                overlap_height_ratio=current_overlap,
                overlap_width_ratio=current_overlap,
                postprocess_type="NMS",
                postprocess_match_metric="IOS",
                postprocess_match_threshold=0.5,
                postprocess_class_agnostic=True,
                verbose=1
            )
            object_prediction_list = result.object_prediction_list
            use_outlier_filter = True

        # Applies a final class-agnostic NMS pass to refine the results.
        object_prediction_list = apply_class_agnostic_nms(object_prediction_list, iou_threshold=0.6)

        class_counts = Counter()
        class_confidences = defaultdict(list)

        # Calculates the median area of all detections to use for outlier filtering.
        all_areas = []
        for pred in object_prediction_list:
            if pred.score.value >= confidence_threshold:
                bbox = pred.bbox
                area = (bbox.maxx - bbox.minx) * (bbox.maxy - bbox.miny)
                all_areas.append(area)

        median_area = np.median(all_areas) if all_areas else 0

        # Iterates through predictions to filter outliers and draw annotations.
        for prediction in object_prediction_list:
            if prediction.score.value < confidence_threshold:
                continue

            bbox = prediction.bbox
            box_area = (bbox.maxx - bbox.minx) * (bbox.maxy - bbox.miny)

            # Applies an outlier filter to remove unusually large detections, which are often false positives in field images.
            if use_outlier_filter and median_area > 0:
                coverage_ratio = box_area / img_area
                # Skip if box covers >5% of image.
                if coverage_ratio > 0.05:
                    continue
                # Skip if box is >15x median area.
                if box_area > (median_area * 15):
                    continue

            # Aggregates statistics for the final summary.
            class_name = prediction.category.name
            score = prediction.score.value
            class_counts[class_name] += 1
            class_confidences[class_name].append(score)

            # Draws the bounding box rectangle onto the image.
            color_hex = color_map.get(class_name, "#ffffff")
            color_bgr = hex_to_bgr(color_hex)
            x_min, y_min = int(bbox.minx), int(bbox.miny)
            x_max, y_max = int(bbox.maxx), int(bbox.maxy)
            cv2.rectangle(image_cv, (x_min, y_min), (x_max, y_max), color_bgr, 2)

        # Visualization and Reporting

        # Converts the OpenCV (BGR) image to RGB for Matplotlib display.
        img_rgb = cv2.cvtColor(image_cv, cv2.COLOR_BGR2RGB)
        fig_w_in, fig_h_in = target_width / dpi, target_height / dpi

        # Creates a two-panel figure: one for the image, one for the legend.
        fig, (ax_image, ax_legend) = plt.subplots(
            1, 2,
            figsize=(fig_w_in, fig_h_in),
            dpi=dpi,
            gridspec_kw={'width_ratios': [3, 1], 'wspace': 0.05}
        )

        ax_image.imshow(img_rgb)
        ax_image.axis('off')
        ax_legend.axis('off')

        # Constructs the legend handles from the aggregated detection data.
        legend_handles = []
        for class_key in lifecycle_order:
            if class_key in class_counts:
                count = class_counts[class_key]
                scores = class_confidences[class_key]
                avg_score = sum(scores) / len(scores) if scores else 0

                label_text = f"{format_class_name(class_key)}: {count} — {avg_score * 100:.2f}%"
                patch = mpatches.Patch(color=color_map.get(class_key, "#000"), label=label_text)
                legend_handles.append(patch)

        # Renders the legend on the right-hand panel.
        if legend_handles:
            ax_legend.legend(
                handles=legend_handles,
                loc='center',
                title="Detection Summary",
                fontsize=24,
                title_fontsize=30,
                frameon=False,
                labelspacing=0.8
            )

        plt.tight_layout()
        plt.show()

        # Prints a final textual summary to the console.
        print("\n" + "-" * 70)
        print(f"Image report: {filename}")
        print("-" * 70)
        for class_key in lifecycle_order:
             if class_key in class_counts:
                scores = class_confidences[class_key]
                avg = sum(scores)/len(scores)
                print(f"{format_class_name(class_key):<15} | Count: {class_counts[class_key]:<5} | Avg Conf: {avg*100:.2f}%")
        print("-" * 70)

# Executes the main analysis function.
run_image_analysis()

----------------------------------------------------------------------
Image report: 02440.jpg
----------------------------------------------------------------------
Adult           | Count: 4     | Avg Conf: 73.27%
----------------------------------------------------------------------