v0.3
v0.3 release notes:
New features:
- Multiple Input Channels Support: Now supporting up to 8 channels for diverse audio setup configurations.
- Data Augmentation: Introducing data augmentation techniques for more robust gesture recognition. This includes applying pitch shift and time stretching to enhance the training set using existing data.
- Prediction Accuracy Threshold: Introducing the prediction accuracy threshold setting which can filter predicted results below a certain accuracy threshold.
- Spectrum Components Settings: Users can now modify the number of frequency components in the spectrogram, allowing for more detailed feature extraction and potentially improving model accuracy.
- OSC Receive Capability: Includes the ability to receive OSC messages to control some parameters on the fly. The receive port is
1123
- Validation Module: The new validation module enables users to test the accuracy of their trained models using saved samples and provides metrics such as gesture-specific accuracy and average accuracy.
- Gesture-Audio Mapping Playback: Enhanced interaction with the ability to map recognised gestures to specific audio playback.
- Machine Learning Model Improvements: Improved the training process of the ML model based on custom gesture recordings and settings.
- Customisable MIDI Channel: Users can now customise the MIDI output channel.
- User Interface Enhancements: The main interface has been refined for better user experience and accessibility.
Bug Fixes
- Various bug fixes and performance improvements.
Full Changelog: v0.25...v0.3