This project implements a real-time face tracking system using MediaPipe, OpenCV, and Dynamixel servos. A camera detects a face, calculates its offset from the center of the frame, and moves two Dynamixel motors to keep the face centered through synchronized position control.
Features:
Real-time face detection using MediaPipe
Smooth pan-tilt tracking with Dynamixel Protocol 2.0
High-speed communication at 4 Mbps
PID-tuned servo response
Safe angle limits and synchronized read/write operations
Tech Stack: Python OpenCV MediaPipe Dynamixel SDK
Overview: The main script (camera.py) handles the full tracking loop: capture frames, detect face, compute offset, convert pixel displacement to servo angle, and send updated goal positions. The repository also includes modules for YOLO/ONNX detection, CUDA-accelerated processing, and Kalman filtering for more advanced tracking options.