Re-implement Full Resolution Image Compression with Recurrent Neural Networks

To train Compression Model (Encoder, Binarizer, Decoder)

CUDA_VISIBLE_DEVICES=0 python train.py --bs 128 --lr 0.0005 --out_dir 'output/' --name 'fullres-compress' --data_path 'CIFAR10-images/train' --save_record False --iterations 16

To train Entropy Coding (Still in Progress)

CUDA_VISIBLE_DEVICES=0 python pixel_code.py

For test arithmetic coding (Still in Progress)

python main.py

To download Datasets

CIFAR-10 Directly

git clone https://github.com/YoongiKim/CIFAR-10-images.git

ImageNet-32x32 (https://image-net.org/download-images.php)

Setup Conda Environment

conda create --name fullres python=3.10
pip install -r requirements.txt

You can just use the yml environment.

conda env create -f /path/to/environment.yml