saw_mill_knot_detection/train_yolov6.py

"""
Train YOLOv6 for knot detection (MIT license - free for commercial use).

YOLOv6 is from Meituan, optimized for deployment on edge devices.

Usage:
    python train_yolov6.py --dataset-dir dataset_prepared --model yolov6n --epochs 100
"""

import argparse
from pathlib import Path
import subprocess
import sys


def train_yolov6(
    dataset_dir: Path,
    output_dir: Path,
    model_name: str = "yolov6n",
    epochs: int = 100,
    batch_size: int = 8,
    img_size: int = 640,
    learning_rate: float = 1e-2,
):
    """
    Train YOLOv6 model.
    
    Args:
        dataset_dir: Path to dataset with train/valid/test splits
        output_dir: Where to save checkpoints
        model_name: One of ['yolov6n', 'yolov6s', 'yolov6m', 'yolov6l']
        epochs: Number of training epochs
        batch_size: Batch size
        img_size: Input image size
        learning_rate: Learning rate
    """
    # Install YOLOv6 if not already installed
    try:
        import yolov6
    except ImportError:
        print("Installing YOLOv6...")
        subprocess.check_call([
            sys.executable, "-m", "pip", "install",
            "git+https://github.com/meituan/YOLOv6.git"
        ])
    
    # Validate dataset structure
    train_dir = dataset_dir / "train"
    valid_dir = dataset_dir / "valid"
    
    if not train_dir.exists() or not valid_dir.exists():
        raise ValueError(f"Dataset must have train/ and valid/ directories")
    
    train_ann = train_dir / "_annotations.coco.json"
    valid_ann = valid_dir / "_annotations.coco.json"
    
    if not train_ann.exists():
        raise ValueError(f"Missing {train_ann}")
    if not valid_ann.exists():
        raise ValueError(f"Missing {valid_ann}")
    
    # Create output directory
    output_dir.mkdir(parents=True, exist_ok=True)
    
    # Create data config file for YOLOv6
    data_yaml = output_dir / "data.yaml"
    with data_yaml.open("w") as f:
        f.write(f"""train: {train_dir.absolute()}
val: {valid_dir.absolute()}

nc: 10
names: ['Live knot', 'Dead knot', 'Knot with crack', 'Crack', 'Resin', 'Marrow', 'Quartzity', 'Knot missing', 'Blue stain', 'Overgrown']
""")
    
    print(f"\n{'='*60}")
    print(f"Training YOLOv6-{model_name}")
    print(f"Dataset: {dataset_dir}")
    print(f"Output: {output_dir}")
    print(f"{'='*60}\n")
    
    # Map model names
    model_map = {
        "yolov6n": "yolov6n",
        "yolov6s": "yolov6s",
        "yolov6m": "yolov6m",
        "yolov6l": "yolov6l",
    }
    
    if model_name not in model_map:
        raise ValueError(f"Model must be one of {list(model_map.keys())}")
    
    # Build training command
    yolov6_dir = Path(sys.executable).parent.parent / "YOLOv6"
    train_script = yolov6_dir / "tools/train.py"
    
    cmd = [
        sys.executable,
        str(train_script),
        "--batch", str(batch_size),
        "--conf", str(yolov6_dir / f"configs/{model_name}.py"),
        "--data", str(data_yaml),
        "--epochs", str(epochs),
        "--device", "0",
        "--name", "yolov6_training",
        "--output-dir", str(output_dir),
    ]
    
    print(f"Running: {' '.join(cmd)}\n")
    
    result = subprocess.run(cmd)
    
    if result.returncode == 0:
        print(f"\n{'='*60}")
        print(f"✓ Training complete!")
        print(f"Weights saved in: {output_dir}/yolov6_training")
        print(f"{'='*60}\n")
    else:
        print(f"\n❌ Training failed with exit code {result.returncode}")
    
    return result.returncode == 0


def main():
    parser = argparse.ArgumentParser(description="Train YOLOv6 for knot detection")
    parser.add_argument(
        "--dataset-dir",
        type=Path,
        required=True,
        help="Path to dataset directory with train/valid/test splits"
    )
    parser.add_argument(
        "--output-dir",
        type=Path,
        default=Path("runs/yolov6_training"),
        help="Output directory for checkpoints and logs"
    )
    parser.add_argument(
        "--model",
        type=str,
        choices=["yolov6n", "yolov6s", "yolov6m", "yolov6l"],
        default="yolov6n",
        help="YOLOv6 model variant (n=smallest/fastest, l=largest/most accurate)"
    )
    parser.add_argument("--epochs", type=int, default=100, help="Number of epochs")
    parser.add_argument("--batch-size", type=int, default=8, help="Batch size")
    parser.add_argument("--img-size", type=int, default=640, help="Input image size")
    parser.add_argument("--lr", type=float, default=1e-2, help="Learning rate")
    
    args = parser.parse_args()
    
    train_yolov6(
        dataset_dir=args.dataset_dir,
        output_dir=args.output_dir,
        model_name=args.model,
        epochs=args.epochs,
        batch_size=args.batch_size,
        img_size=args.img_size,
        learning_rate=args.lr,
    )


if __name__ == "__main__":
    main()