Breaking changes:
- now imports require only one eyegestures reference
import eyegestures
Features:
- Adding new v3 model,
- v3 model comes with experimental saccades detection,
- v3 face stabilization (user's head is tracked to counter act its movements),
- v3 face scaling (when user's head is moving forward or backward head landmarks will scaled),
- v3 do not use v1 nor v2,
- added test targets to v3 demo
- low pass filter for keypoints in v3
Fixes:
- More equal distribution of calibration data (now it has 30 data points per calibration point)