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.