Developed an AI-powered PPE detection system using Python, Ultralytics YOLOv8, PyTorch, and OpenCV to detect 17 safety equipment classes from images, videos, and live webcam streams. Implemented IoU-based worker-PPE association with context-aware compliance logic, achieving 63.24% precision, 48.65% recall, 51.68% mAP@50, and 31.87% mAP@50–95.