Realtime ONNX Handtracking

I created a real-time hand tracking system using ONNX models and OpenCV that detects a user's hand and tracks 21 landmarks, color-coded by depth. The project taught me ONNX inference, preprocessing, and coordinate mapping, and I overcame challenges like stabilizing detection at different distances by designing a dynamic bounding box system. Unlike most solutions that rely on MediaPipe libraries or Python bridges, this project uses standalone ONNX models in a single Python script, which simplifies the pipeline significantly. It's well suited for integration into Unity using the Sentis package, and my next goal is to bring it into an AR app without external dependencies.