🐛 Bug Fixes #

• Fixed potential crashes in Android release builds by adding consumer ProGuard rules. This ensures MediaPipe's essential classes are not stripped by R8. (46a978a)

🛠️ Maintenance #

• Updated camera plugin dependency to version 0.11.3.

📝 Documentation #

• Added additional examples for plugin usage in README.md.

🐛 Bug Fixes #

• Fixed a java.lang.UnsatisfiedLinkError crash on 32-bit (armeabi-v7a) devices by updating the native MediaPipe tasks-vision dependency to 0.10.26.1. (Fixes #1)

✨ Features #

• Configurable Options: Added the ability to configure the hand landmarker with the following options:
  • numHands: The maximum number of hands to detect.
  • minHandDetectionConfidence: The minimum confidence score for hand detection to be considered successful.
  • delegate: The delegate to use for inference, allowing for selection between CPU and GPU.
  • see e1dff3e

💥 Breaking Changes #

• Synchronous API: The plugin's core methods are now synchronous to improve performance. This affects how you create, use, and dispose of the plugin.
  • HandLandmarkerPlugin.create() no longer returns a Future.
  • HandLandmarkerPlugin.dispose() is now synchronous.
  • The detect() method is now a synchronous, blocking call. You must manage how frequently you call it to avoid blocking the UI thread.

✨ Features & Performance #

• Native Image Processing (BREAKING): Rearchitected the plugin to perform all YUV image conversion natively in Kotlin. This eliminates the Dart background isolate and significantly reduces data transfer overhead for much lower latency. (347f5f1)
• GPU Acceleration: Enabled the MediaPipe GPU delegate by default to accelerate model inference, resulting in smoother real-time performance. (4749d8c)

• Initial release of the project.