Added #

• HEIC Image Format Support: Native HEIC/HEIF image decoding on iOS and Android
  • Automatic platform-specific native decoding using platform channels
  • Efficient decoding suitable for large datasets (thousands of images)
  • No manual setup required - platform code automatically included
  • Support for iOS 11+ and Android 9+ (API 28+)
• Image Format Detector: New utility class ImageFormatDetector for automatic image format detection from bytes
  • Supports JPEG, PNG, BMP, WebP, HEIC, and HEIF formats
  • Detects format using file signatures (magic numbers)
  • Exported in main library for public use

Improved #

• Text Recognition Empty Text Handling: Enhanced ML Kit text recognition to properly handle empty/no text cases
  • Filters out empty and whitespace-only text blocks
  • Returns "No text detected" message instead of empty text with high confidence
  • Better user feedback when no text is found in images
• Face Detection Empty Face Handling: Enhanced ML Kit face detection to properly handle empty/no face cases
  • Checks for empty face lists and zero face counts
  • Returns "No face detected" message instead of empty results
  • Consistent behavior with text recognition for better user feedback
• Android Package Structure: Restructured as proper Flutter plugin with native platform code
  • Platform code automatically registered - no manual integration needed
  • Proper plugin configuration in pubspec.yaml
  • Native iOS and Android implementations included in package

Changed #

• Android Package Name: Updated to com.github.dhia_bechattaoui.flutter_ml_helper following standard reverse-domain notation
• Supported Image Formats: Added 'heic' and 'heif' to MLConstants.supportedImageFormats list

Fixed #

• Text Recognition Results: Fixed issue where empty text was shown with high confidence scores
• Android Build Configuration: Added required namespace declaration in android/build.gradle for newer Android Gradle Plugin versions

Technical Details #

• HEIC images are decoded using native iOS ImageIO and Android BitmapFactory APIs
• Decoded HEIC images are converted to PNG format internally for compatibility with Dart image package
• Platform-specific native code located in android/ and ios/ directories
• Automatic plugin registration via Flutter's plugin system

Fixed #

• CI Dependency Resolution: Fixed dependency conflicts with test, mockito, and build_runner packages to ensure compatibility with Flutter SDK test_api constraints

Changed #

• Dev Dependencies: Updated dev dependencies for better CI compatibility:
  • test: ^1.0.0 (downgraded from ^1.26.2 to resolve test_api conflicts)
  • mockito: ^5.0.0 (downgraded from ^5.5.0)
  • build_runner: ^2.0.0 (downgraded from ^2.7.1)

Added #

• ImageNet Class Labels Support: Automatic loading of ImageNet-1K class labels (1000 classes) from PyTorch Hub
• Smart Model Detection: Automatic detection of ImageNet models (MobileNet, ResNet, Inception, etc.) by output shape and model name
• Flexible Image Normalization: Support for both [0, 1] and [-1, 1] normalization ranges in preprocessImageForML()
• ImageNetLabels Public API: Exported ImageNetLabels utility class for accessing ImageNet class labels
• Enhanced TFLite Inference: Automatic softmax application for logits-to-probabilities conversion
• HTTP Package: Added http: ^1.0.0 dependency for fetching ImageNet labels from URL

Improved #

• Better Classification Results: Fixed normalization to use [-1, 1] range by default for MobileNet models
• Automatic Label Loading: ImageNet labels are automatically loaded when ImageNet models are detected
• Fallback Support: Falls back to hardcoded labels if network request fails
• Example App: Fixed BuildContext async usage issues in example application
• Code Quality: Improved linting compliance and error handling

Technical Details #

• ImageNet labels are cached in memory after first load for fast subsequent access
• Labels are loaded asynchronously and non-blocking
• Supports both quantized and float32 TFLite models
• Automatic detection based on 1000-class output shape or model name keywords

Dependencies #

• http: ^1.0.0 (new)

Added #

• Initial release of Flutter ML Helper package
• Support for TensorFlow Lite integration
• Support for Google ML Kit integration
• Cross-platform compatibility (iOS, Android, Web, Windows, macOS, Linux)
• WASM compatibility for web platform
• Core ML helper utilities and classes
• Image processing capabilities
• Permission handling for device access
• Path management for model files

Technical Features #

• Flutter SDK 3.32.0+ compatibility
• Dart SDK 3.8.0+ compatibility
• Comprehensive test coverage (>90%)
• Pana score: 160/160
• Linting and code quality tools
• Build runner support for code generation

Platform Support #

• ✅ iOS
• ✅ Android
• ✅ Web (with WASM support)
• ✅ Windows
• ✅ macOS
• ✅ Linux

Dependencies #

• tflite_flutter: ^0.10.4
• google_ml_kit: ^0.16.3
• image: ^4.1.7
• path_provider: ^2.1.2
• permission_handler: ^11.3.1