Pose Estimation & Human Motion

What was the Problem?

Physical therapy and posture correction usually require in-person coaching or expensive sensor arrays.

How I Solved It

Designed a real-time posture detection system that runs efficiently on standard mobile hardware using YOLO and TFLite.

Effort Reduced

Allowed users to receive instant posture feedback at home without needing a human coach to monitor their exercises continuously.

Profit / Impact Achieved

Created a scalable software solution with minimal operational costs, resulting in a patent-filed technology.

Difficulties Faced

Mobile devices lack the processing power to run standard heavy pose-estimation models in real-time without draining the battery.

How I Overcame It

Optimized the model architecture and utilized TensorFlow Lite to achieve real-time inference (30+ FPS) on mobile chips without sacrificing accuracy.