Object Detection with YOLO

Build a custom object detection system using YOLOv8.

Project details:
- Objects to detect: [LIST OF CLASSES]
- Dataset format: [COCO/YOLO/Pascal VOC]
- Number of images: [DATASET SIZE]
- Deployment target: [EDGE/SERVER/MOBILE]

Requirements:
1. Data preparation:
   - Format conversion
   - Annotation validation
   - Train/val/test split
2. Data augmentation:
   - Mosaic, mixup
   - Color/geometric transforms
3. Model configuration:
   - Model size selection (n/s/m/l/x)
   - Custom architecture modifications
4. Training:
   - Transfer learning from COCO weights
   - Learning rate scheduling
   - Multi-GPU training
5. Evaluation:
   - mAP calculation
   - Per-class performance
   - Inference speed benchmarking
6. Post-processing:
   - NMS tuning
   - Confidence thresholds
7. Export and deployment:
   - ONNX, TensorRT, CoreML
   - Real-time inference optimization

Use Ultralytics YOLOv8 library.