MSU Compressed Video Quality Dataset (CVQAD) and results verification for Video Quality Metrics Benchmark, introduced in the NeurIPS 2022 paper «Video compression dataset and benchmark of learning-based video-quality metrics»

Check out more detailed charts and leaderboards on the MSU Video Quality Metrics Benchmark page.

To reproduce paper results just run

final_scripts_for_correlation.ipynb

You are provided with metrics scores on the open part (named CVQAD) of our benchmark dataset. We also share the connections between each reference video and the categories into which they have been assigned.

Citation

If you use our benchmark results, CVQA Dataset, or this code for your research, please cite our paper.

@inproceedings{
    NEURIPS2022_59ac9f01,
    author = {Antsiferova, Anastasia and Lavrushkin, Sergey and Smirnov, Maksim and Gushchin, Aleksandr and Vatolin, Dmitriy and Kulikov, Dmitriy},
    booktitle = {Advances in Neural Information Processing Systems},
    editor = {S. Koyejo and S. Mohamed and A. Agarwal and D. Belgrave and K. Cho and A. Oh},
    pages = {13814--13825},
    publisher = {Curran Associates, Inc.},
    title = {Video compression dataset and benchmark of learning-based video-quality metrics},
    url = {https://proceedings.neurips.cc/paper_files/paper/2022/file/59ac9f01ea2f701310f3d42037546e4a-Paper-Datasets_and_Benchmarks.pdf},
    volume = {35},
    year = {2022}
}

Contact

We would highly appreciate any suggestions and ideas on how to improve our benchmark.

Maksim Smirnov, [email protected]