MSU Video Upscalers Benchmark 2021
The most comprehensive comparison of video super resolution (VSR) algorithms by subjective quality
Video group head: Dr. Dmitriy Vatolin
Measurements, analysis:
Nikolai Karetin,
Ivan Molodetskikh
Ivan Molodetskikh
Our benchmark will help you to select the best upscaling approach for your workflow. Check the “Methodology” section to learn the details.
Everyone is welcome to participate! Run your favorite super-resolution method on our small test video and send us the result to see how well it performs. Check the “Submitting” section to learn the details.
What’s new
- October 11, 2021: Alpha-version Release
Key features of the Benchmark
- Subjective comparison with more than 4300 valid participants
- Visual comparison for all 40 algorithms on all 15 clips
- You can check how visually good each state-of-the-art method is
- Check how upscalers behave in the most practical upscale use cases with great variety of content types and distortions
- Find the best method that produces great results on various content types
- Special calculation of objective metrics with ¼ pixel precision robust to object borders displacement
- Metric loss map visualizations
If you are specifically interested in detail-preserving upscaling, check the MSU Video Super-Resolution Benchmark.
If you are specifically interested in compressed video upscaling, check the MSU Super-Resolution for Video Compression Benchmark.
4× Results
Check the “Participants” section to learn the details about used algorithms.
FPS was calculated running on NVIDIA Titan Xp, if not stated otherwise.