pose_detection_tflite 2.0.1

Pose & landmark detection using on-device TFLite models.

2.0.1

• Migrate iOS CocoaPods -> Swift Package Manager

2.0.0

Breaking changes:

• Remove all deprecated image package-based APIs
  • PoseDetector.detect(List<int>) — use detect(cv.Mat) instead
  • PoseDetector.detectOnImage(img.Image) — use detect(cv.Mat) instead
  • PoseDetector.imageDecoderOverride setter — no longer needed
  • PoseLandmarkModelRunner.run(img.Image) — use run(cv.Mat) instead
  • YoloV8PersonDetector.detectOnImage(img.Image) — use detect(cv.Mat) instead
  • ImageUtils.letterbox(), letterbox256(), imageToNHWC4D() — use NativeImageUtils equivalents
  • NativeImageUtils.imageToMat() — use cv.imdecode directly
• Remove image package dependency

1.3.3

• Upgrade flutter_litert -> 0.1.12

1.3.2

• Windows: remove bundled .dll files, as they are no longer needed as of flutter_litert 0.1.4

1.3.1

• Swift Package Manager support

1.3.0

• Deprecate all img.Image-based APIs in favor of cv.Mat alternatives
• PoseDetector.detect() and detectOnImage() → use detectOnMat()
• PoseLandmarkModelRunner.run() → use runOnMat()
• YoloV8PersonDetector.detectOnImage() → use detectOnMat()
• ImageUtils.letterbox(), letterbox256(), imageToNHWC4D() → use NativeImageUtils equivalents
• NativeImageUtils.imageToMat() → use cv.imdecode or pass cv.Mat directly
• The image package dependency will be removed in 2.0.0

1.2.2

• Fix Android build: bump tflite_flutter_custom to 1.2.5 (fixes undefined symbol TfLiteIntArrayCreate linker error)
• Add missing MainActivity.kt for Android example app

1.2.1

• Improved auto .dylib bundling on MacOS

1.2.0

• Add native OpenCV preprocessing for 5-15x faster image processing (SIMD-accelerated)
• New useNativePreprocessing parameter on PoseDetector (default: true)
• New detectOnMat() method on YoloV8PersonDetector for direct cv.Mat input
• New runOnMat() method on PoseLandmarkModelRunner for direct cv.Mat input
• Add opencv_dart dependency for native image operations
• Refactored detection code with shared post-processing helpers

1.1.1

• Cache yolo input buffer and pass flat tensor to avoid reshape overhead
• Update tflite_flutter_custom to 1.1.0
• 100% test coverage

1.1.0

• Parallel landmark detection in multi person photos, faster processing
• Improve error handling
• Add integration tests

1.0.1

• Improved Dartdoc coverage

1.0.0

This version contains breaking changes:

• Breaking API changes: PoseDetector uses constructor params (no PoseOptions), returns Pose/BoundingBox, and detection methods are async
• YOLO and BlazePose run via isolates to avoid blocking
• Expanded DartDoc coverage
• poseLandmarkConnections for drawing skeletons.
• Example app improvements, sample images added.

0.0.3

• Removed unused dependency path_provider

0.0.2

• Improved documentation

0.0.1

• Initial release
• Person detection using YOLOv8
• Pose landmark detection with MediaPipe Pose
• Support for lite, full, and heavy models
• Box-only and full landmark detection modes