face_detection_tflite 4.1.0

face_detection_tflite: ^4.1.0 copied to clipboard

• Native image processing with opencv_dart for ~2x performance improvement via SIMD acceleration
  • detectFaces() now uses OpenCV internally
  • New detectFacesFromMat() method for camera streams (avoids repeated encode/decode overhead)
• XNNPACK delegate enabled by default for 2-5x CPU speedup (use PerformanceConfig.disabled to opt out)
• Benchmark tests

Breaking changes:

• Replace math.Point<double> type references with Point
• Change face.mesh.isEmpty to face.mesh == null
• Access mesh points via face.mesh?.points[i] or face.mesh?[i]
• Replace face.irises → face.eyes
• Replace IrisPair → EyePair
• Replace iris.center → eye.irisCenter
• Replace iris.contour → eye.irisContour

Improvements:

• Performance and speed improvements
  • Optimize bilinear sampling with direct buffer access, 20-40% speed improvement
  • Fast-path frame registration
  • Parallel iris refinement
  • Isolate-based image-to-tensor conversion.
• Improved test suite, added integration tests

• EyePair class and eye mesh landmarks (71 points per eye)
• Add eyeContour getter for accessing visible eyelid outline (first 15 of 71 points)
• Add eyeLandmarkConnections constant for rendering connected eyelid outline
• Add kMaxEyeLandmark constant defining eyeball contour point count

• Guard iris ROI size and fall back when eye crop collapses

• Add frame registration fast path to reduce transfers
• Parallel iris refinement for multi-face
• Cache input tensor buffers in FaceDetection, FaceLandmark, and IrisLandmark

• Performance improvement: Optimize full mode by reusing mesh and iris landmarks
• Add pub.dev score and version to README

This version contains breaking changes.

• Remove deprecated bboxCorners and landmarksMap. The new BoundingBox and FaceLandmarks class should be used instead.
• Rename bbox to boundingBox.

• FaceLandmarks class instead of Map
• Simplified process of accessing individual landmarks

• BoundingBox class
• Add clarification to README about dart:math requirement for Point

• Bundle Point from dart:math with library
• Improved dartdocs

• Fix bug in example related to overlay rect scaling

• Added missing dartdoc for entry point lib/face_detection_tflite.dart and RectF constructor in lib/src/types_and_consts.dart

• New IrisPair class
• Add structured iris types, rename raw iris points
• Update example, README usage examples

• Add Dartdocs for AlignedFace, AlignedRoi, DecodedBox, DecodedRgb
• Minor clarifications to existing getDetectionsWithIrisCenters and RectF Dartdocs
• Improved examples in README

• Swift Package Manager support

• Update tflite_flutter_custom to 1.0.3, equivalent to tflite_flutter 0.12.1.
• Improved dartdocs

This version contains breaking changes.

• detectFaces now returns List
• Public types renamed/privatized (FaceIndex is now FaceLandmarkType, RectF/Detection/ AlignedFace/AlignedRoi now internal).
• Landmark maps now keyed by FaceLandmarkType
• Full dartdoc coverage

• Update tflite_flutter_custom to 1.0.1, equivalent to tflite_flutter 0.12.0.
• Unit tests
• Performance optimization(s) by enabling parallel inferences in images with multiple faces

• Three detection modes: fast, standard & full. Enables faster inferences when the full detection set is not needed.

• Improved error handling
• Added samples, improved documentation
• Improved example (see example tab on pub.dev)

• Provide end-user with pre-normalized, image-space coords.
• Improved readme/documentation & public API as a whole,
• Removed obsolete methods, change Offset objects to Point.

• Fix bug where IrisLandmark inferences would fail in an Isolate

• Moved heavy operations to Isolates to avoid UI clank/lag

• Refresh iOS/Android example project files to avoid stale tool warnings.

• Tweak analysis/lints config to match latest Flutter stable.

• Minor bug fixes & improvements
• Clarifications in the README

• Add iOS and Android via dartPluginClass
• Keep native plugin on desktop (macOS/Windows/Linux) so CMake still bundles TFLite C libs.
• Note: iOS release builds may require Xcode “Strip Style = Non-Global Symbols”; test on device (not simulator).
• Note: Android requires minSdk 26 (handled by the app).

• Initial public release of face_detection_tflite.
• Includes TFLite face detection + landmarks models and platform shims.
• Adds prebuilt libtensorflowlite_c for macOS/Windows/Linux.