Deep Wyner Stereo Image Compression (DWSIC) [Paper]

"Ballé2018" and "DSIN" refer to Variational image compression with a scale hyperprior, Deep Image Compression using Decoder Side Information, respectively. "Ours" refers to our proposed model using "Ballé2017" model, which refers to End-to-end Optimized Image Compression, as the baseline.

Citation

Deep Stereo Image Compression with Decoder Side Information using Wyner Common Information [Paper]

@misc{DWSIC,
      title={Deep Stereo Image Compression with Decoder Side Information using Wyner Common Information}, 
      author={Nitish Mital and Ezgi Ozyilkan and Ali Garjani and Deniz Gunduz},
      year={2021},
      eprint={2106.11723},
      archivePrefix={arXiv},
      primaryClass={eess.IV}
}

Abstract

We present a novel deep neural network (DNN) architecture for compressing an image when a correlated image is available as side information only at the decoder. This problem is known as distributed source coding (DSC) in information theory. In particular, we consider a pair of stereo images, which generally have high correlation with each other due to overlapping fields of view, and assume that one image of the pair is to be compressed and transmitted, while the other image is available only at the decoder. In the proposed architecture, the encoder maps the input image to a latent space, quantizes the latent representation, and compresses it using entropy coding. The decoder is trained to extract the Wyner's common information between the input image and the correlated image from the latter. The received latent representation and the locally generated common information are passed through a decoder network to obtain an enhanced reconstruction of the input image. The common information provides a succinct representation of the relevant information at the receiver. We train and demonstrate the effectiveness of the proposed approach on the KITTI dataset of stereo image pairs. Our results show that the proposed architecture is capable of exploiting the decoder-only side information, and outperforms previous work on stereo image compression with decoder side information.

Usage

Clone

Clone this repository and enter the directory using the commands below:

git clone https://github.com/ipc-lab/DWSIC.git
cd DWSIC/

Requirements

Python 3.7.3 is recommended.

Install the required packages with:

pip install -r requirements.txt

If you're having issues with installing PyTorch compatible with your CUDA version, we strongly suggest you refer to related documentation page.

Dataset

The dataset used for experiments is KITTI Stereo. You can download the necessary image pairs from KITTI 2012 and KITTI 2015. After obtaining data_stereo_flow_multiview.zip and data_scene_flow_multiview.zip, run the following commands:

unzip data_stereo_flow_multiview.zip # KITTI 2012
mkdir data_stereo_flow_multiview
mv training data_stereo_flow_multiview
mv testing data_stereo_flow_multiview

unzip data_scene_flow_multiview.zip # KITTI 2015
mkdir data_scene_flow_multiview
mv training data_scene_flow_multiview
mv testing data_scene_flow_multiview

Weights

Pre-trained models on the KITTI Stereo setup mentioned in the paper for "Ballé2017", "Ballé2018", "Ours + Ballé2017" and "Ours + Ballé2018" models, trained either wrt. MSE or MS-SSIM reconstruction loss, can be downloaded from this link.

If desired, download the desired weights and put them in pretrained_weights folder and update the configs/config.yaml parameters accordingly.

Getting Started

To use the code, please run:

python main.py

Please be aware that the initial console output (including the ongoing test results) might take a while to be printed.

By default, this code uses the configurations in configs/config.yaml. You can either change this configuration file or create a new yaml file and use that for running. For this case, please run:

python main.py --config=path/to/new/config/file

You can also use Custom Configuration Notebook.ipynb notebook to create your custom configuration and run the model based on that.

Configurations

• Dataset:

dataset_path: '.'
resize: [128, 256]

dataset_path shows the path to data_paths directory that contains every image and its pair path. The resize value selects the width, and the height dimensions that each image will be resized to.

• Model:

baseline_model: 'bls17' # can be bmshj18 for Variational image compression with a scale hyperprior by Ballé, et al.
                          # or bls17 for End-to-end Optimized Image Compression by Ballé, et al.
use_side_info: True # if True then the modified version of baseline model for distributed compression is used.
num_filters: 192 # number of filters used in the baseline model network
cuda: True
load_weight: False
weight_path: './pretrained_weights/balle17+ours_MS-SSIM_lambda3e-05.pt' # weight path for loading the weight
# note that we provide some pretrained weights, accessible from the link provided in README.md, under the title "Weights"

baseline_model selects the compression model. The accepted models for this parameter are 'bmshj18' for Variational image compression with a scale hyperprior and 'bls17' for End-to-end Optimized Image Compression. If use_side_info is set as True, then the baseline model is modified using our proposed method for using side information for compressing. If load_weight is True, then in model initialization, the weight saved in weight_path is loaded to the model.

• Training

train: True
epochs: 50000
train_batch_size: 1
lr: 0.0001
lambda: 0.00003 # the lambda value in rate-distortion equation
alpha: 1
beta: 1
distortion_loss: 'MS-SSIM' # can be MS-SSIM or MSE. selects the method by which the distortion is calculated during training
verbose_period: 50 # non-positive value indicates no verbose

For training set train to be True. lambda shows the lambda value in the rate-distortion equation and alpha and beta correspond to the handles on the reconstruction of the correlated image and amount of common information extracted from the decoder-only side information, respectively. distortion_loss selects the distortion evaluating method. Its accepted values are MS-SSIM for the ms-ssim method or MSE for mean squared error. verbose_period: 50 indicates that every 50 epochs print the results of the validation dataset.

• Weight parameters

save_weights: True
save_output_path: './outputs' # path where results and weights will be saved
experiment_name: 'bls17_with_side_info_MS-SSIM_lambda:3e-05'

If you wish to save the model weights after training set save_weights True. save_output_path shows the directory path where the model weights are saved. For the weights, in save_output_path a weight folder will be created, and the weights will be saved there with the name according to experiment_name.

• Test:

test: True
save_image: True
experiment_name: 'bls17_with_side_info_lambda:0.1'

If you wish to test the model and save the results set test to True. If save_image is set to True then a results folder will be created, and the reconstructed images will be saved in save_output_path/results during testing, with the results named according to experiment_name.

Inference

In order to (only) carry out inference, please open configs/config.yaml and change the relevant lines as follows:

resize: [128, 256] # we used this crop size for our inference
dataset_path: '.'
train: False
load_weight: True
test: True
save_output_path: './inference' 
save_image: True 

Download the desired weights and put them in pretrained_weights folder and put the dataset folder in the root .

Based on the weight you chose, specify the weight name in configs/config.yaml:

weight_path: './pretrained_weights/...'  # load a specified pre-trained weight
experiment_name: '...' # a handle for the saved results of the inference

Also, change baseline_model and use_side_info parameters in configs/config.yaml accordingly. For example, for the balle2017+ours weights, these parameters should be:

baseline_model: 'bls17'
use_side_info: True

After running the code using the command below, the results will be saved in inference folder.

python main.py

Results on KITTI Stereo

In Figure 8. (a) and (b), we used MSE and MS-SSIM distortion functions for training the models, respectively. The values of the parameter lambda for the DNN models, and the values of the quality parameter for BPG and JPEG2000, used to obtain the plotted points are given in Results.md.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Authors/Contributors

• Ezgi Ozyilkan* (github: ezgimez)
• Ali Garjani* (github: garjania)
• Nitish Mital* (github: nitishmital)
• Deniz Gunduz (github: dgunduz)

*These authors contributed equally to this work.