Vinted Start-up

A mobile app that helps Vinted sellers take great product photos effortlessly: the camera detects the garment in real time, gives live instructions, and once everything is aligned it takes the picture by itself and crops it to the product.

Repository structure

Folder	Contents
VisionApp/	The React Native (TypeScript) app plus the native C++/OpenCV plugin
VisionApp/cpp/	Vision pipeline: ProductAnalyzer, SensorFusion, SegmentationEngine
VisionApp/src/	UI, hooks (auto-capture, photo workflow, alignment), frame processor
VisionApp/android/, VisionApp/ios/	Native projects

Full technical documentation (pipeline, thresholds, design decisions, caveats) lives in VisionApp/README.md.

Features

• Real-time detection of the product, processed in C++ with OpenCV directly on camera frames (JSI, zero copies).
• Live guidance: one message at a time (e.g. "Center the garment", "Hold the phone straighter").
• Sensor fusion (accelerometer + gyroscope) for tilt and flat-lay / on-hanger mode.
• Auto-capture + auto-crop once the framing is correct and steady.
• Vinted photo flow: Front → Back → Label → Defects/Details.

Quick start

Requirements: Node.js, JDK + Android SDK/NDK (Android) or Xcode + CocoaPods (iOS).

cd VisionApp
npm install

# iOS
cd ios && bundle install && bundle exec pod install && cd ..
npx react-native run-ios

# Android
npx react-native run-android

Start Metro separately with npm start. Run tests with npm test.

Tech stack

React Native 0.86 · TypeScript · VisionCamera · Reanimated · C++20 · OpenCV 4.10 · LiteRT for ML segmentation (in preparation).

Status

Under active development. Tested on an Android emulator; the build is configured for arm64-v8a and x86_64. Real-device benchmarks have not been recorded yet.