EsakaK / LSSVC

Official implementation of LSSVC: A Learned Spatially Scalable Video Coding Scheme.
MIT License
5 stars 1 forks source link

LSSVC: A Learned Spatially Scalable Video Coding Scheme

Official Pytorch implementation for LSSVC: A Learned Spatially Scalable Video Coding Scheme

Prerequisites

Build the project

We only provide the test script with real bitstream writing. Please build the C++ code to test with actual bitstream writing.

On Linux

sudo apt-get install cmake g++
cd src
mkdir build
cd build
conda activate $YOUR_PY36_ENV_NAME
cmake ../cpp -DCMAKE_BUILD_TYPE=Release
make -j

Comparing with other method

Pretrained models

We provide our pretrained models:

IntraSS is a degraded version of LSSVC without interframe references.

Command Lines

We provide the command lines of the encoder and decoder of HM-18.0, VTM-21.2, and SHM. The VTM-21.2 is utilized for both simulcast coding and two-layer scalable coding in our experiments.

simulcast coding

For simulcast encoding and decoding, it involves performing two separate single-layer encoding and decoding processes. Here, we summarize the command lines used for single-layer coding.

two-layer coding

Citation

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

@ARTICLE{10521480,
  author={Bian, Yifan and Sheng, Xihua and Li, Li and Liu, Dong},
  journal={IEEE Transactions on Image Processing}, 
  title={LSSVC: A Learned Spatially Scalable Video Coding Scheme}, 
  year={2024},
  volume={33},
  number={},
  pages={3314-3327},
  keywords={Video coding;Encoding;Image coding;Standards;Scalability;Spatial resolution;Static VAr compensators;Learned video coding;spatial scalability;scalable video coding;contextual MV encoder-decoder;hybrid temporal-layer context mining;interlayer prior},
  doi={10.1109/TIP.2024.3395025}}