Object Tracking in MP with OpenVINO Inference

This project aims to integrate bytetrack algorithm into a Mediapipe graph using OpenVINO inference calculators. The bytetrack graph will be deployed on OpenVINO model server to perform efficient object tracking on a video input. The deliverables of this project include a bytetrack graph in OpenVINO mediapipe fork with custom calculators, YOLOv10 model and bytetrack graph packaged in OpenVINO model server, and a Python client for the bytetrack graph to perform bytetrack algorithm on video stream. The final phase of this project includes evaluation of bytetrack on metrics MOTA, IDF1, and HOTA. The goal of this project is to bring the bytetrack algorithm to OpenVINO model server to leverage its efficiency, scalability, and robustness for production-grade inference.