I developed a real-time computer vision system for human pose estimation using MediaPipe BlazePose. The application detects 33 body landmarks and reconstructs a live skeleton using OpenCV. It also computes joint angles, tracks movement trajectories, and classifies human activities such as walking, running, sitting, and standing. A PyQt5 GUI was built for professional visualization, with multi-threaded processing ensuring real-time performance above 30 FPS. The system supports data export in CSV and JSON formats for further analysis.