Awesome

SADN: Learned Light Field Image Compression with Spatial-Angular Decorrelation + QVRF: A Quantization-error-aware variable rate framework for learned image compression

Pytorch implementation of the paper "SADN: Learned Light Field Image Compression with Spatial-Angular Decorrelation" with variable rate techniche of "QVRF: A Quantization-error-aware variable rate framework for learned image compression"

Related links

• CompressAI: https://github.com/InterDigitalInc/CompressAI
• SADN：https://github.com/VincentChandelier/SADN
• QVRF：https://github.com/bytedance/QRAF

Installation

Install CompressAI and the packages required for development.

conda create -n SADN+QVRF python=3.9
conda activate FPIcompress
pip install compressai==1.1.5
pip install ptflops
pip install einops
pip install tensorboardX

Trainning

stage 1

Train a high-rate fixed-rate model.

python train.py -d dataset --N 48 --angRes 13 --n_blocks 1 -e 100 -lr 1e-4 -n 20  --lambda 3e-3 --batch-size 8  --test-batch-size 8 --aux-learning-rate 1e-3 --patch-size 832 832 --cuda --save --seed 1926 --clip_max_norm 1.0 --gpu-id  2,3 --savepath   ./checkpoint --stage 1 --ste 0 --loadFromPretrainedSinglemodel 0  

stage 2

Load the last checkpoint of stage 1 or load from the Fixed-rate model of SADN and begin to train the variabal rate model in stage 2.

python train.py -d dataset --N 48 --angRes 13 --n_blocks 1 -e 100 -lr 1e-4 -n 20  --lambda 3e-3 --batch-size 8  --test-batch-size 8 --aux-learning-rate 1e-3 --patch-size 832 832 --cuda --save --seed 1926 --clip_max_norm 1.0 --gpu-id  2,3 --savepath  ./Noisecheckpoint --stage 2 --ste 0 --loadFromPretrainedSinglemodel 1 --checkpoint checkpoint.pth.tar --pretrained

stage 3

Load the last checkpoint of stage 2 and begin to train the final variabal rate model in stage 3.

python train.py -d dataset --N 48 --angRes 13 --n_blocks 1 -e 100 -lr 1e-6 -n 20  --lambda 3e-3 --batch-size 8  --test-batch-size 32 --aux-learning-rate 1e-3 --patch-size 832 832 --cuda --save --seed 1926 --clip_max_norm 1.0 --gpu-id  2,3 --savepath  ./STEcheckpoint --stage 3 --ste 1 --loadFromPretrainedSinglemodel 0 --checkpoint ./Noisecheckpoint/checkpoint_best_loss_82.pth.tar --pretrained

Test

Full-resolution test images

python3 Inference.py --dataset ./dataset/Fulltest --s 2 --output_path SADNSTE -p ./Proposed_STE__28_checkpoint.pth.tar --patch 832 --factormode 0 --factor 0.1

If you have any problem, please contact me: tkd20@mails.tsinghua.edu.cn

If you think it is useful for your reseach, please cite our paper.

Citation

@inproceedings{tong2022sadn,
  title={SADN: learned light field image compression with spatial-angular decorrelation},
  author={Tong, Kedeng and Jin, Xin and Wang, Chen and Jiang, Fan},
  booktitle={ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={1870--1874},
  year={2022},
  organization={IEEE}
}

@article{tong2023qvrf,
  title={QVRF: A Quantization-error-aware Variable Rate Framework for Learned Image Compression},
  author={Tong, Kedeng and Wu, Yaojun and Li, Yue and Zhang, Kai and Zhang, Li and Jin, Xin},
  journal={arXiv preprint arXiv:2303.05744},
  year={2023}
}