Home

Awesome

CANF-VC: Conditional Augmented Normalizing Flows for Video Compression

Update (09.11.23): CANF-VC++: Enhancing Conditional Augmented Normalizing Flows for Video Compression with Advanced Techniques

Update (08.30.22): CANF-VC with Error Propagation Aware Training Strategy

Project Installation

  1. Prepare PyTorch 1.4.0 environment and correspond torchvision
  2. Run sh install.sh
  3. (Only needed for CANF-VC*) Install libbpg: https://github.com/mirrorer/libbpg 3.1 Configure path to libbpg as libbpg_path in dataloader.py
  4. Download model weights & prepare testing data
  5. Start evaluation: action=test/compress/decompress

Model Weight

Dataset

Examples

Full Commands

Citation

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

@article{canfvc,
  title={CANF-VC: Conditional Augmented Normalizing Flows for Video Compression},
  author={Ho, Yung-Han and Chang, Chih-Peng and Chen, Peng-Yu and Gnutti, Alessandro and Peng, Wen-Hsiao},
  journal={European Conference on Computer Vision},
  year={2022}
}