fixin - Web
Real-Time Computer Vision Interface (Desktop / Browser)
Developed a browser-based version of fixin that runs the full AI/ML diagnostic pipeline client-side via TensorFlow.js or TFLite— no install, no native build, no server round-trip for inference. Designed primarily for desktop and browser-based access, it shares the same backend API as the iOS app and extends the platform beyond Apple hardware.
Key features:
-
Real-time object detection (COCO-SSD via TensorFlow.js/TFLite) running entirely in the browser
-
Live camera pipeline with bounding boxes and confidence overlays for interactive user feedback
-
Validation system cross-referencing real-time detections against server-side ML model outputs to improve accuracy
-
Frame throttling (30fps) and optimized client-side processing for consistent performance
-
Full integration with the fixin backend API: diagnostics, history tracking, repair workflows, and marketplace