All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Add Files to dataset from manifest.
- Add Detected samples to dataset.
- Samples smart trimming.
- Noise and noised sample management.
- CRNN and CNN Model Architectures.
- Models layer additions.
- Models dropout and Gaussian Layers.
- LMFE Features.
- Inference optimizations.
New Software architecture. Projects created with older version will no longer work. Please create a new project.
- Project now support multiple datasets, features and model architectures.
- Complete code overhaul.
- Dependencies including tensorflow latest version.
- Module are now presented as widgets within in the navigation interface.
- All module redevelopped using a new _Module pattern.
- All module now accept multiple profiles
- Dataset, features, model and trained_model classes.
- Navigation interface
- Added Datasets Management Module
- Export and import existing datasets
- Sample's sampling rate are now not irrelevant as they are automaticly resampled to the target sampling rate.
- FeatToFile and FileToFeat functions.
- Implemented homebrew MFCC feature extraction.
- Remove and/or delete samples (with manifest) function in the evaluate Module.
- Removed large chunk of now unused code.
- Added unroll option at model creation to unroll the GRU layer in order to export TF Lite format.
- Added tensorflow lite export for unrolled model.
- Add changelog
- Fix #1. Division by zero