liveness_edge_flutter 0.3.2

Private, offline face-liveness verification for Flutter using MediaPipe and ONNX Runtime.

0.3.2

• Process sequential camera frames with MediaPipe video mode on Android and iOS, enabling inter-frame landmark tracking without changing liveness rules or the Flutter API.
• Preserve the display-oriented BGRA and bi-planar YUV coordinates delivered by camera_avfoundation so iOS face-position guidance does not swap axes.
• Match Android horizontal face-position guidance to the mirrored front-camera preview without changing head-turn challenge directions.

0.3.1

• Detect unreliable or out-of-frame eye and mouth landmarks on Android and iOS, and show unified guidance when required facial regions are not clearly visible.
• Tolerate sub-500 ms detector dropouts, then restart the complete verification when the face or required landmarks remain obscured and continuity is lost.
• Harden same-face verification by requiring a sustained, consistent identity mismatch before restarting and stable recovery frames before continuing.

0.3.0

• Add LivenessEdgeTheme for screen colors and camera-guide appearance.
• Move guideline and supporting text styles into LivenessEdgeTheme; retain the screen-level properties as deprecated compatibility aliases.
• Add oval, circular, and rounded-rectangle camera shapes.
• Add headerBuilder with read-only LivenessEdgeViewState presentation data.
• Improve open-mouth recognition on iOS with a normalized lip-gap fallback when the MediaPipe jawOpen blendshape is too conservative.
• Show move-farther guidance before the face reaches the frame boundary, improving alignment feedback on iOS front-camera frames.

0.2.1

• Require three stable, frontal frames with both eyes fully open before saving the final verification image.
• Prevent fast iOS devices from exhausting the analyzed-frame safety cap well before the configured session timeout.
• Tolerate brief face-landmarker dropouts during active challenges while still enforcing same-face identity when detection resumes.
• Preserve more landmark detail when downscaling iOS BGRA camera frames without changing the Android YUV/NV21 conversion path.
• Normalize front-camera yaw on iOS so left and right head-turn instructions match the user's physical direction.

0.2.0

• Randomly select three or four active challenges per session from blink, one randomly directed head turn, smile, and open mouth.
• Validate smile and open-mouth transitions with MediaPipe blendshapes.

0.1.0

• Initial offline Android and iOS implementation.
• MediaPipe face landmarks, blink blendshapes, and head-turn estimation.
• MiniFASNetV2 passive anti-spoofing through ONNX Runtime.
• Dart challenge state machine and ready-to-use camera screen.
• Custom guideline text, styles, and retry button builder.
• Complete setup, usage, security, and API documentation.