D2
badminton-vision – Real-Time On-Device Computer Vision for Badminton
Independent Developer · Jun 2025 – present
Fully on-device Chrome MV3 extension for real-time badminton match analysis, with browser-based pose, racket, and shuttle detection; actively rebuilding toward its first stable release.
Technologies
TypeScriptChrome Extension (Manifest V3)TensorFlow.jsONNX Runtime WebYOLO-World
Contributions
- Built a Chrome MV3 extension that detects pose, racket, and shuttle on badminton match footage played on YouTube, running entirely client-side with no server round-trip.
- Implemented the default real-time detection pipeline with TensorFlow.js and ONNX Runtime Web.
- Evaluated an experimental zero-shot, open-vocabulary object detector (YOLO-World) alongside the default detector.
- Applied full product-development discipline: a written PRD, UX wireframing and review, an automated test suite, and a CI pipeline.
- Public on GitHub (Jin-HoMLee/badminton-vision); still in active development, not yet a finished v1.
Outcome
In-browser, on-device computer vision extension for badminton match analysis, under active, ongoing development toward a first stable release.