CPGA

The dataset and code of the paper "CPGA: Coding Priors-Guided Aggregation Network for Compressed Video Quality Enhancement".

Requirements

CUDA==11.6 Python==3.7 Pytorch==1.13

1.1 Environment

conda create -n cpga python=3.7 -y && conda activate cpga

git clone --depth=1 https://github.com/VQE-CPGA/CPGA && cd VQE-CPGA/CPGA/

# given CUDA 11.6
python -m pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116

python -m pip install tqdm lmdb pyyaml opencv-python scikit-image

1.2 DCNv2

cd ops/dcn/
bash build.sh

Check if DCNv2 work (optional)

python simple_check.py

1.3 VCP dataset

Download raw and compressed videos

Please check BaiduPan,Code [qix5].

Edit YML

You need to edit option_CPGA_vcp_#_QP#.yml file.

Generate LMDB

The LMDB generation for speeding up IO during training.

python create_vcp.py --opt_path option_CPGA_vcp_#_QP#.yml

Finally, the VCP dataset root will be sym-linked to the folder ./data/ automatically.

1.4 Test dataset

We use the JCT-VC testing dataset in JCT-VC. Download raw and compressed videos BaiduPan,Code [qix5].

Train

python train_CPGA.py --opt_path ./config/option_CPGA_vcp_LDB_22.yml

Test

python test_CPGA.py --opt_path ./config/option_CPGA_vcp_LDB_22.yml

Citation

If this repository is helpful to your research, please cite our paper:

@inproceedings{zhu2024cpga,
  title={CPGA: Coding Priors-Guided Aggregation Network for Compressed Video Quality Enhancement},
  author={Zhu, Qiang and Hao, Jinhua and Ding, Yukang and Liu, Yu and Mo, Qiao and Sun, Ming and Zhou, Chao and Zhu, Shuyuan},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition}
}
@article{zhu2024deep,
  title={Deep Compressed Video Super-Resolution With Guidance of Coding Priors},
  author={Qiang Zhu, Feiyu Chen, Yu Liu, Shuyuan Zhu, Bing Zeng},
  journal={ IEEE Transactions on Broadcasting }
}
@article{zhu2024compressed,
  title={Compressed Video Quality Enhancement with Temporal Group Alignment and Fusion},
  author={Qiang, Zhu and Yajun, Qiu and Yu, Liu and Shuyuan, Zhu and Bing, Zeng},
  journal={IEEE Signal Processing Letters}
}
@inproceedings{mo2025oapt,
  title={OAPT: Offset-Aware Partition Transformer for Double JPEG Artifacts Removal},
  author={Mo, Qiao and Ding, Yukang and Hao, Jinhua and Zhu, Qiang and Sun, Ming and Zhou, Chao and Chen, Feiyu and Zhu, Shuyuan},
  booktitle={European Conference on Computer Vision}
}

Related Works

We also released some compressed video quality enhancement models, e.g., STDF, RFDA, CF-STIF, and STDR.

Our project is built on the STDF. If there are some problems with the implementation, please refer to STDF. We adopt Apache License v2.0. For other licenses, please refer to DCNv2.