Appearance
Step 0 Pre-Check: Project Type Detection
Detected Project Type
Primary Type: Client-Side Computer Vision & Touchless Human-Computer Interaction (HCI) Web Application (React 18 + Google MediaPipe Vision Tasks + WebAssembly + WebGL GPU Delegate) Sub-Category: Real-Time Optical Landmark Classification, Hands-Free Video Player for Kitchen & Workshop Environments, Zero-Server Privacy Architecture.
Evidence
- Computer Vision Framework:
@mediapipe/tasks-vision0.10.14 (package.json,src/services/GestureManager.js),FilesetResolver.forVisionTasks,GestureRecognizer.createFromOptions. - Inference Mode: Local WebAssembly runtime (
float16/1/gesture_recognizer.task) running client-side with GPU delegate (delegate: "GPU"). - Frontend Architecture: React 18.3.1, Vite 5.4.11, Tailwind CSS 3.4.15,
lucide-react0.441.0. - Hardware Interface:
navigator.mediaDevices.getUserMedia640x480 30fps optical webcam stream.
Documentation Track
- Computer Vision / Frontend HCI Track: Deep dive into in-browser Wasm execution, tensor confidence thresholds, frame-rate stabilization loops, gesture debounce heuristics, and zero-latency client-side interaction.