Labo Compressie 2018

a project for B-KUL-YI1328

Usage

tested on ubuntu 18.04

clone this repo git clone https://github.com/JosseVanDelm/2018_labo_compressie_JosseVanDelm.git
install all dependencies for keras docker container e.g. nvidia-docker (cf. KerasREADME.md)
run make notebook (note token output on terminal)
go to http://localhost:8888/ in a webbrowser, enter token from previous step here
upload both .ipynb-files of this directory to the notebook.
open the Convolutional Autoencoder (Training).ipynb file, this is the training-script. Start the training process. (takes about 30 minutes on CPU, and about 4 on GPU, depending on your system)
open the Encoder and Decoder.ipynb file. This file uses the trained weights from the training phase to encode and decode any given image. You will notice however that this last script doesn't always provide the wanted outcome :(

Name		Name	Last commit message	Last commit date
Latest commit History 9 Commits
presentation		presentation
Autoencoder (with sparsity constraint).ipynb		Autoencoder (with sparsity constraint).ipynb
Autoencoder Keras.ipynb		Autoencoder Keras.ipynb
Convolutional Autoencoder (Training).ipynb		Convolutional Autoencoder (Training).ipynb
Convolutional autoencoder.ipynb		Convolutional autoencoder.ipynb
Dockerfile		Dockerfile
Encoder and Decoder (separate).ipynb		Encoder and Decoder (separate).ipynb
KerasREADME.md		KerasREADME.md
Makefile		Makefile
README.md		README.md
theanorc		theanorc