Cassini Hackathon 2024 - YOLO Object Detection

• Military drone detection
• Error detection in images/video

Project Structure

├── train.py              # Main training script
├── *.pt files            # PyTorch model weights (best/last checkpoints)
├── drone_mil.v1i.yolov11/    # Military drone detection dataset
├── FindError.v4i.yolov11/    # Error detection dataset
├── runs/                 # Training results and predictions
├── video/               # Test videos (DJI drone footage)
└── yolo11n_ncnn_model/  # NCNN optimized model

Features

• Drone Detection: Military applications using YOLO v11
• Error Detection: Automated error identification in visual data
• Multiple Model Variants: Different model sizes (nano, small, medium)
• Video Processing: Real-time detection on drone footage

Models

• yolo11n.pt - YOLO v11 Nano (fastest)
• yolo11s.pt - YOLO v11 Small (balanced)
• Various trained models with different configurations

Usage

Dataset Information

Military Drone Dataset

• Source: drone_mil.v1i.yolov11
• Classes: Military drone detection
• Format: YOLO format with train/valid/test splits

Error Detection Dataset

• Source: FindError.v4i.yolov11
• Classes: Various error types
• Format: YOLO format with train/valid/test splits

Results

• runs/detect/ - Prediction outputs
• benchmarks.log - Performance metrics

Note

Hackathon Context