flutter_hand_gesture 2.1.1

A Flutter package for real-time hand gesture recognition using on-device ML. Supports custom gestures, landmark detection, and stream-based API. No internet required.

2.1.1

• Docs: update README to document all v2 features — Tamil/ISL sign presets, SignRecognizer, gesture sequences, and the action mapper. No code changes.

2.1.0

• SignRecognizer — disambiguates signs that share a finger shape by also using hand orientation and motion. Thumbs-up (nalla) vs thumbs-down (ketta) and the wagging index (illai) are now told apart.
• HandOrientation and SignMotion enums, plus a SignPattern interface implemented by TamilSign and IslAlphabetSign.
• TamilSign / IslAlphabetSign gain optional orientation and motion fields (default any / none) — no breaking changes.
• Tamil demo now uses SignRecognizer with a zero-debounce stream so motion is detectable.

2.0.0

• Tamil/ISL sign language preset (15 signs + 5 vowel fingerspelling).
• Gesture sequence recognition with stream binding and progress tracking.
• Sequence progress painter widget.
• Gesture → Action mapper with accessibility and presentation presets.
• No breaking changes to the v0.2.0 API.

0.1.0

• Initial release.
• Real on-device hand-landmark detection via the hand_detection package (MediaPipe/TFLite, bundled models, background isolate) — fully offline.
• 10 built-in gestures: fist, open palm, thumbs up/down, peace, pointing, rock, call me, OK, vulcan. The common six use MediaPipe's trained gesture recognizer; the rest use rule-based landmark geometry.
• 21-point hand landmarks with helper geometry (HandLandmarks).
• Stream-based API (gestureStream, landmarkStream).
• GestureCameraView plug-and-play widget with skeleton overlay.
• CustomGestureTrainer with a KNN classifier and disk persistence.
• GestureConfig for confidence, hand count, target gestures, debounce, etc.
• Frame throttling to keep CPU/GPU load and heat manageable.
• Requires Android minSdkVersion 24.