Add comprehensive test suite and feature parity documentation

.github/workflows/tests.yml

@@ -0,0 +1,71 @@
+name: Tests
+
+on:
+  push:
+    branches: [main, master]
+  pull_request:
+    branches: [main, master]
+
+jobs:
+  typescript-tests:
+    name: TypeScript Tests
+    runs-on: self-hosted
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: '20'
+          cache: 'npm'
+
+      - name: Install dependencies
+        run: npm ci --legacy-peer-deps
+
+      - name: Run TypeScript tests
+        run: npm test
+
+      - name: Run TypeScript tests with coverage
+        run: npm run test:coverage
+        continue-on-error: true
+
+  android-tests:
+    name: Android Unit Tests
+    runs-on: self-hosted
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Setup Java
+        uses: actions/setup-java@v4
+        with:
+          distribution: 'temurin'
+          java-version: '17'
+
+      - name: Setup Android SDK
+        uses: android-actions/setup-android@v3
+
+      - name: Run Android unit tests
+        working-directory: android
+        run: ./gradlew test --no-daemon
+
+  # Uncomment when MacBook Air is set up as a self-hosted runner
+  # ios-tests:
+  #   name: iOS Unit Tests
+  #   runs-on: self-hosted-macos  # Update with your macOS runner label
+  #
+  #   steps:
+  #     - name: Checkout
+  #       uses: actions/checkout@v4
+  #
+  #     - name: Run iOS tests
+  #       working-directory: ios
+  #       run: |
+  #         xcodebuild test \
+  #           -scheme VisionML \
+  #           -destination 'platform=iOS Simulator,name=iPhone 15' \
+  #           -only-testing:VisionMLTests \
+  #           | xcpretty

FEATURE_PARITY.md

@@ -0,0 +1,245 @@
+# Feature Parity: iOS vs Android
+
+This document tracks the feature parity between iOS and Android implementations of react-native-vision-ml.
+
+## Summary
+
+| Category | iOS | Android | Status |
+|----------|-----|---------|--------|
+| Core ONNX Inference | ✅ | ✅ | **Full Parity** |
+| Video Analysis | ✅ | ✅ | **Full Parity** |
+| Vision/ML Kit | ✅ | ⚠️ | **Partial** |
+| Sensitive Content Analysis | ✅ | ❌ | **iOS Only** |
+| Live Activity | ✅ | ⚠️ | **Partial** |
+
+---
+
+## Detailed Feature Comparison
+
+### Core ONNX Inference ✅ Full Parity
+
+| Feature | iOS | Android | Notes |
+|---------|-----|---------|-------|
+| Model Loading | ✅ CoreML | ✅ NNAPI | Both use hardware acceleration |
+| Image Decode | ✅ CGImage | ✅ Bitmap | HEIC/JPEG/PNG supported |
+| NCHW Conversion | ✅ | ✅ | Identical tensor format |
+| YOLO v8 Parsing | ✅ | ✅ | Same output format |
+| NMS Algorithm | ✅ | ✅ | Identical implementation |
+| Multiple Detectors | ✅ | ✅ | Concurrent instances supported |
+| Resource Cleanup | ✅ | ✅ | dispose/disposeAll methods |
+
+**Performance:**
+- iOS: ~120-150ms per image (CoreML acceleration)
+- Android: ~120-150ms per image (NNAPI acceleration)
+
+### Video Analysis ✅ Full Parity
+
+All 5 video scan modes are implemented identically:
+
+| Mode | iOS | Android | Description |
+|------|-----|---------|-------------|
+| `quick_check` | ✅ | ✅ | 3 frames: start, middle, end |
+| `sampled` | ✅ | ✅ | Regular interval sampling |
+| `thorough` | ✅ | ✅ | Human detection + ONNX on human frames |
+| `binary_search` | ✅ | ✅ | Adaptive region expansion |
+| `full_short_circuit` | ✅ | ✅ | Progressive scan with early exit |
+
+**Frame Extraction:**
+- iOS: AVAssetImageGenerator
+- Android: MediaMetadataRetriever.getFrameAtTime()
+
+### Vision Framework / ML Kit ⚠️ Partial Parity
+
+| Feature | iOS (Vision) | Android (ML Kit) | Status |
+|---------|--------------|------------------|--------|
+| Animal Detection | ✅ VNRecognizeAnimalsRequest | ✅ Image Labeling | **Parity** |
+| Human Pose | ✅ VNDetectHumanBodyPoseRequest | ✅ Pose Detection | **Parity** |
+| Face Detection | ✅ VNDetectFaceRectanglesRequest | ✅ Face Detection | **Parity** |
+| Text Recognition | ✅ VNRecognizeTextRequest | ✅ Text Recognition | **Parity** |
+| Scene Classification | ✅ VNClassifyImageRequest | ❌ Not Available | **Gap** |
+| Rectangle Detection | ✅ VNDetectRectanglesRequest | ❌ Not Available | **Gap** |
+
+#### Gap: Scene Classification
+- **iOS**: VNClassifyImageRequest provides ~1000 scene categories
+- **Android**: ML Kit doesn't have an equivalent API
+- **Workaround**: Would require custom TensorFlow Lite model
+
+#### Gap: Rectangle Detection
+- **iOS**: VNDetectRectanglesRequest for screenshot/document detection
+- **Android**: Not implemented
+- **Impact**: `likelyScreenshot` field always false on Android
+
+### Sensitive Content Analysis ❌ iOS Only
+
+| Feature | iOS | Android | Notes |
+|---------|-----|---------|-------|
+| Single Image SCA | ✅ iOS 17+ | ❌ | Apple proprietary |
+| Batch Image SCA | ✅ iOS 17+ | ❌ | Apple proprietary |
+| Video SCA | ✅ iOS 17+ | ❌ | Apple proprietary |
+| Batch Video SCA | ✅ iOS 17+ | ❌ | Apple proprietary |
+| SCA Settings | ✅ | ❌ | Apple proprietary |
+
+**Why No Android Support:**
+Apple's Sensitive Content Analysis framework is a proprietary iOS 17+ feature with no Android equivalent. Google does not provide a similar on-device content analysis API.
+
+**Workaround for Android:**
+Use the ONNX-based detection which works on both platforms and provides detailed bounding boxes and confidence scores.
+
+### Live Activity ⚠️ Partial Parity
+
+| Feature | iOS | Android | Notes |
+|---------|-----|---------|-------|
+| Start Activity | ✅ Dynamic Island | ⚠️ Stub | Foreground service infrastructure ready |
+| Update Progress | ✅ Real-time | ⚠️ Stub | Notification UI not implemented |
+| End Activity | ✅ Final status | ⚠️ Stub | |
+
+**Android Implementation Status:**
+- Foreground service infrastructure is in place
+- Notification progress UI is stubbed (always returns success but no visible UI)
+- Full implementation requires custom notification layout
+
+---
+
+## API Compatibility
+
+### Methods with Full Parity
+
+```typescript
+// These work identically on both platforms
+createDetector(modelPath, classLabels, inputSize)
+detect(detectorId, imageUri, options)
+disposeDetector(detectorId)
+disposeAllDetectors()
+analyzeVideo(detectorId, assetId, options)
+quickCheckVideo(detectorId, assetId, threshold)
+analyzeAnimals(assetId)
+analyzeHumanPose(assetId)
+```
+
+### Methods with Platform Differences
+
+```typescript
+// Works but may return partial results on Android
+analyzeComprehensive(assetId)
+// - Android: scenes=[], rectangles=0, likelyScreenshot=false
+
+// iOS 17+ only - Android returns {available: false}
+getSensitiveContentAnalysisStatus()
+analyzeSensitiveContent(assetId)
+batchAnalyzeSensitiveContent(assetIds)
+analyzeVideoSensitiveContent(assetId)
+batchAnalyzeVideosSensitiveContent(assetIds)
+
+// iOS 16.1+ only - Android returns stub responses
+isLiveActivityAvailable()
+startVideoScanActivity(name, duration, mode)
+updateVideoScanActivity(progress, phase, count, frames)
+endVideoScanActivity(count, frames, isNSFW)
+```
+
+---
+
+## Recommended Usage Patterns
+
+### Cross-Platform Detection
+
+```typescript
+// Works on both iOS and Android
+const detector = await createDetector(modelPath, labels, 320);
+const result = await detect(detector.detectorId, imageUri, {
+  confidenceThreshold: 0.6,
+  iouThreshold: 0.45,
+});
+```
+
+### Platform-Specific Features
+
+```typescript
+import { Platform } from 'react-native';
+
+// Use SCA on iOS 17+, fall back to ONNX on Android
+if (Platform.OS === 'ios') {
+  const status = await getSensitiveContentAnalysisStatus();
+  if (status.available && status.enabled) {
+    const result = await analyzeSensitiveContent(assetId);
+    if (result.isSensitive) {
+      // Run ONNX for detailed bounding boxes
+      const detailed = await detect(detectorId, imageUri);
+    }
+  }
+} else {
+  // Android: Use ONNX directly
+  const result = await detect(detectorId, imageUri);
+}
+```
+
+### Comprehensive Analysis
+
+```typescript
+const analysis = await analyzeComprehensive(assetId);
+
+// Handle platform differences
+const hasSceneData = Platform.OS === 'ios' && analysis.scenes?.length > 0;
+const isScreenshot = Platform.OS === 'ios' && analysis.likelyScreenshot;
+```
+
+---
+
+## Test Coverage
+
+### TypeScript Tests (Jest)
+- API integration tests
+- NMS algorithm unit tests
+- YOLO parser unit tests
+- Platform parity verification
+
+Run with: `npm test`
+
+### Android Tests (JUnit)
+- NMS algorithm tests
+- YOLO parser tests
+- Coordinate transformation tests
+
+Run with: `./gradlew test` in android/
+
+### iOS Tests (XCTest)
+- NMS algorithm tests
+- YOLO parser tests
+- Coordinate transformation tests
+
+Run with: Xcode Test Navigator
+
+---
+
+## Roadmap
+
+### Planned Improvements
+
+1. **Android Live Activity** (P1)
+   - Implement notification progress UI
+   - Match iOS Dynamic Island UX
+
+2. **Android Scene Classification** (P2)
+   - Research custom TensorFlow Lite models
+   - Consider Google Cloud Vision API fallback
+
+3. **Android Rectangle Detection** (P3)
+   - Implement ML Kit custom model integration
+   - Or use OpenCV-based detection
+
+### Not Planned
+
+- **Android SCA**: No equivalent API exists; use ONNX detection instead
+
+---
+
+## Version History
+
+| Version | Changes |
+|---------|---------|
+| 1.0.5 | Video SCA fixes |
+| 1.0.4 | Added video SCA support |
+| 1.0.3 | Added batch SCA support |
+| 1.0.2 | Full Android implementation |
+| 1.0.1 | Initial Android stubs |
+| 1.0.0 | Initial iOS-only release |