HessamKaveh/wildlife-object-detection

Object detection on African wildlife images using fine-tuned YOLOv8 # African Wildlife Object Detection (YOLOv8) Fine-tuned YOLOv8n on the African Wildlife dataset to detect 4 species: buffalo, elephant, rhino, and zebra. ## Pipeline 1. Load pretrained YOLOv8n (COCO weights) 2. Fine-tune on African Wildlife dataset (auto-downloaded via Ultralytics) 3. Train for 30 epochs with early stopping (patience=10) 4. Evaluate: mAP50, mAP50-95, confusion matrix, PR curve 5. Run inference on held-out test images ## Dataset African Wildlife Dataset (auto-downloaded on first run via `african-wildlife.yaml`) ## Installation ```bash pip install -r requirements.txt ``` ## Usage ```bash cd src python train.py python evaluate.py ``` ## Results See `results/wildlife_yolo/results.png` for training curves and `results/predictions/` for sample detections. ## Author Hessam Kaveh — Research Fellow, Italian Institute of Technology