Keras extended Utilities for MLSP feature learning
The project contains tools for easier development with Keras/Tensorflow. It was developed for AVA-MLSP feature learning. The code in applications.py and model_helper.py pertains more to MLSP feature learning, wheras the remaining tools are general purpose.
Some of the key components of each file:
H5Helper: Manage named data sets in HDF5 files, for us in Keras generators.
pretty: Pretty-print dictionary type objects.
ShortNameBuilder: Utility for building short (file) names that contain multiple parameters.
ImageAugmenter: Create custom image augmentation functions for training Keras models.
read_image_batch: utility functions for manipulating images.
ModelHelper: Wrapper class that simplifies default usage of Keras for regression models.
model_inceptionresnet_multigap: Model definitions for extracting MLSP narrow features
model_inceptionresnet_pooled: Model definitions for extracting MLSP wide features
DataGeneratorHDF5: Keras generators for on-disk images, and HDF5 stored features/images
You can find more information in the Python Docstrings.