Machine Learning - Object Detection, Traffic Analyzer
Engineered a high-speed object detection pipeline to monitor multi-lane traffic from CCTV feeds. Implemented YOLO-World for zero-shot detection, allowing the system to identify new vehicle types without retraining. Integrated ByteTrack for robust multi-object tracking across frames. Optimized model using TensorRT for edge deployment, achieving 45 FPS on an NVIDIA Jetson.