video-autoencoder - Implementation of a simple video auto-encoder, that takes two videos (say 240p and 120p) and creates a convolutional + LSTM network that tries to compress the given video. Can be used to generate a new video of a smaller dimension (or even increase resolution, if a better model would be implemented)
variational-autoencoder - Implementation of a standard VAE, a reconstruction CVAE and a class-based CVAE, tested on MNIST
generative-adversian-network - Implementation of the standard GAN, tested on MNIST
neural_wrappers/
dataset_reader/ - Base class for all readers and various readers for known datasets.
models/ - Various models from different papers implemented. May contain additional or missing features from
original articles (just PyTorch for now).
transforms/ - Basic transforms and some built-in transforms for data augmentation (mirror, cropping)
wrappers/ - Main wrapper directory
pytorch/ - Files that implement various features on top of PyTorch framework
callbacks.py - Basic callback class and some built-in callbacks for training (history.txt, model saving)
metrics.py - Basic metrics and some built-in metrics for training (accuracy/loss)
utils.py - Various functions
reader_converters - Various converters from the form the dataset is offered online (usually a big archive of images and labels, textual or not) to the form that is accepted by an implemented reader (under neural_wrappers/readers), which uses the DatasetReader class API (compatible with Keras fit_generator method and NeuralNetworkPyTorch train_generator). Generally, these converters will generate one h5py file that is used in the reader.
test/
Unit tests for all the implemented modules (WIP)
To run tests, go in the tests directory and type 'pytest' in the console. Requires the pytest module to be
installed, which can be done using pip install pytest
The Tidelift Subscription provides access to a continuously curated stream of human-researched and maintainer-verified data on open source packages and their licenses, releases, vulnerabilities, and development practices.