Image Compression with Recurrent Neural Network and Generalized Divisive Normalization

Prerequisites

The implementation of out network is in Python 3.5.6 and PyTorch. We recommended using conda to install the dependencies. First, create a Python 3.5.6 environment. At this moment, my env name is "Fresh_RNN"

git clone https://github.com/khawar512/ImageCompression
cd ImageCompression
conda create -n Fresh_RNN python=3.5.6    
conda activate Fresh_RNN

pip users:

pip3 install -r requirements.txt

Then, install basic dependencies with conda or pip

Requirements

numpy==1.15.2
matplotlib==3.0.0
scipy==1.1.0
torchvision==0.6.1
pillow==5.2.0
torch==1.5.1

If your machine has multiple GPUs, you can select which GPU you want to run on by setting the environment variable,

CUDA_VISIBLE_DEVICES=0

Data Preparation

We first need to prepare the training data. We take approximately 3700 images that generates 9 million patches. The data is from flicker.com. You can download data from and patch generation file from Link. Then generate the 32*32 patches using following script.

Testing has been done on popular Kodak Photo dataset

bash test/download_kodak.sh

Training

Loading data takes time

python train.py -f /path/32x32_images

Download Models

We provide pre-trained models on above datasets for architecture defined in paper: We will soon make more chnages in models and arhitecture. Watch out for changes to this repo. "Note", Image encoding and decoding utilizes GPU, if you have no GPU then please copy all data inside "results" folder. You can easily produce same results. Pre-trained Models

Testing on pre-trained models

Encode and decode with model

bash test/enc_dec.sh

Citation

If you find this code useful for your research, please cite our work

@InProceedings{Islam_2021_CVPR,
    author    = {Islam, Khawar and Dang, L. Minh and Lee, Sujin and Moon, Hyeonjoon},
    title     = {Image Compression With Recurrent Neural Network and Generalized Divisive Normalization},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month     = {June},
    year      = {2021},
    pages     = {1875-1879}
}

Contact

If you find any problem in code and want to ask any question, please send us email khawar512@gmail.com, khawarislam@fuuast.edu.pk

Acknowledgment

This implementation of code is heavily borrows from Biao Zhang