OVQE: Omniscient Network for Compressed Video Quality Enhancement

The PyTorch implementation for the OVQE: Omniscient Network for Compressed Video Quality Enhancement which is accepted by [IEEE TBC].

Task: Video Quality Enhancement / Video Artifact Reduction.

1. Pre-request

1.1. Environment

Suppose that you have installed CUDA 10.1, then:

conda create -n cvlab python=3.7 -y  
conda activate cvlab
git clone --depth=1 https://github.com/pengliuhan/OVQE && cd OVQE/
python -m pip install torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html
python -m pip install tqdm lmdb pyyaml opencv-python scikit-image thop

1.2. DCNv2

Build DCNv2

cd ops/dcn/
bash build.sh

Check if DCNv2 works (optional)

python simple_check.py

The DCNv2 source files here is different from the open-sourced version due to incompatibility. [issue]

1.3. Dataset

Please check here.

1.4. Create LMDB

We now generate LMDB to speed up IO during training.

python create_lmdb_ovqe.py

2. Train

We utilize 2 NVIDIA GeForce RTX 3090 GPUs for training.

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 --master_port=12354 train.py --opt_path option_ovqe.yml

3. Test

Pretrained models can be found here: [GoogleDisk] and [百度网盘]

We utilize 1 NVIDIA GeForce RTX 3090 GPU for testing.

python test_one_video.py

Citation

If you find this project is useful for your research, please cite:

@article{peng2022ovqe,
  title={OVQE: Omniscient Network for Compressed Video Quality Enhancement},
  author={Peng, Liuhan and Hamdulla, Askar and Ye, Mao and Li, Shuai and Wang, Zengbin and Li, Xue},
  journal={IEEE Transactions on Broadcasting},
  year={2022},
  publisher={IEEE}
}

Acknowledgements

This work is based on STDF-Pytoch. Thank RyanXingQL for sharing the codes.

maouestc/OVQE