Live demos

Interactive ML systems, served from self-hosted inference.

The frontend runs on Vercel. Inference is designed to run on the UN1290 through Cloudflare Tunnel, with static fallback results when the server is offline.

Fashion Attribute Recognition

Upload a garment image and inspect four DINOv2 multi-head predictions with confidence bars.

Open demo

Squat Counter

Upload a squat video and review rep count, form degradation, and per-rep movement metrics.

Open demo