hand_detection 2.0.2

hand_detection: ^2.0.2 copied to clipboard

• Update flutter_litert 2.0.5 -> 2.0.6

• Fix Xcode build warnings by declaring PrivacyInfo.xcprivacy as a resource bundle in iOS and macOS podspecs

Breaking: Point now uses double coordinates. BoundingBox.toMap() format changed to corner-based.

• Use shared Point and BoundingBox from flutter_litert 2.0.0
• toPixel() now returns full-precision double coordinates (was truncating to int)
• Remove duplicate NMS implementation, use shared nms() from flutter_litert
• Refactor isolate worker to use IsolateWorkerBase from flutter_litert
• Simplify model classes (PalmDetector, HandLandmarkModel, GestureRecognizer)
• Remove integration tests from unit test suite
• Remove dead test helpers (test_config.dart)

• Update camera_desktop 1.0.1 -> 1.0.3

• Update flutter_litert -> 1.2.0
• Refactor to use flutter_litert shared utilities (InterpreterFactory, InterpreterPool, PerformanceConfig, generateAnchors)

• Update opencv_dart 2.1.0 -> 2.2.1
• Update flutter_litert 1.0.2 -> 1.0.3

First stable release of hand_detection

Pipeline #

• Palm detection — SSD model with rotation-aware bounding boxes
• Hand landmarks — 21-point 3D landmarks with visibility scores
• Gesture recognition — 7 gestures (fist, open palm, pointing up, thumbs down/up, victory, I love you)
• Handedness — Left/right classification

Features #

• Two modes: HandMode.boxes (bounding boxes only) and HandMode.boxesAndLandmarks (full landmarks)
• HandDetectorIsolate for background-thread inference with zero-copy transfer
• Direct cv.Mat input for live camera processing
• XNNPACK hardware acceleration with configurable thread count
• Configurable confidence thresholds and detection limits

Platforms #

• iOS, Android, macOS, Windows, Linux

• Update documentation

• Update flutter_litert to 1.0.1, camera to 0.12.0

• Update flutter_litert to 0.2.2

• Initial release