Official Project Demonstration: Bharat AI-SOC Student Challenge Finalist Entry
Recognition
This project was selected as a finalist in the Bharat AI-SOC Student Challenge, demonstrating the power of Edge AI in creating intuitive, touchless user experiences.
Official Selection List: Bharat AI-SOC Student Challenge
Project Overview
Ishara is a high-performance touchless interface designed for the Bharat AI-SOC Student Challenge. It leverages advanced skeletal tracking to control media players with zero physical contact, optimized specifically for the Raspberry Pi 5.
Technical Innovations
- MediaPipe BlazePalm Integration: Implements 21-point skeletal tracking, replacing noise-sensitive contour methods with robust 3D topology.
- HFSM Optimization: Engineered a Hierarchical Finite State Machine to replace heavy ML classifiers, reducing CPU load to an ultra-lean 16.9%.
- Scale-Invariant Algorithm: Uses centroid normalization, allowing seamless operation at distances up to 2.4 meters.
- Sliding Window Median Filter: Eliminates tracking jitter, maintaining a deterministic latency of 150ms.
- Asynchronous Execution: Utilizes Pynput for real-time HID command emulation to control VLC Media Player.