FaceVision

Face recognition products usually have an uncomfortable slide in the deck: 'and then we upload your face to our server.' FaceVision removes that slide. Detection (YuNet), recognition (SFace, 128-d embeddings) and anti-spoofing (MiniFASNet V2) all run in the browser with ONNX Runtime Web — WebGPU first, WASM fallback. The FastAPI + PostgreSQL backend only ever sees coordinates and vectors. Never an image. Here's the architecture, the models, the numbers (96.9% on 2,200 real LFW pairs) and the part I'd still fix. Built at NextUpgrad Web Solutions; code on GitHub. I also build tools.scult.in — 15 free tools, 1,200+ prompts and a 50,000-skill library, free for anyone, no signup.