ml-compression #

Automatically create compression schemes convolutional neural networks. This is also known as an Auto Encoder.

Description #

Train a neural network to have its output reproduce the exact input. Why is this interesting? You can constrict the quantity of neurons in one of the intermediate layer and use this as a stand-in for the data that was input. This then means that you have effectively compressed this data.

These types of models can help us generate optimal compression schemes.

Setup #

pip install -r requirements.txt

Running #

python3 main.py

Results #

TODO: #

• Restrict pixels instead of just channels in the convolutions