This forum uses cookies
This forum makes use of cookies to store your login information if you are registered, and your last visit if you are not. Cookies are small text documents stored on your computer; the cookies set by this forum can only be used on this website and pose no security risk. Cookies on this forum also track the specific topics you have read and when you last read them. Please confirm whether you accept or reject these cookies being set.

A cookie will be stored in your browser regardless of choice to prevent you being asked this question again. You will be able to change your cookie settings at any time using the link in the footer.

Questions regarding BasicVSR++
#1
I noted that in Hybrid BasicVSR++ is available under 2 different filter tab:

1) Vapoursynth->Frame->Resize

   where the available models are: Vimeo-90K (BI), Vimeo-90K (BD)
   It is not clear to me the difference of these 2 models.

2) Vapoursynth->Other

    where the available models are: NITRE 2021 (1), NITRE 2021 (2), NITRE 2021 (3)
    Even in this case it is not clear to me difference of these 3 models.

Could you explain the usage of these 2 different BasicVSR++ modules.

At the address https://github.com/open-mmlab/mmediting/.../README.md

there is a reference on these models, but nothing that helps in understanding the difference, moreover it seems that all these models are used for the Video Super-Resolution.
Reply
#2
Difference between the models is just that they are trained with different content and thus will have different effects on different content.
NTIRE = New Trends in Image Restoration and Enhancement workshop and challenges on image and video processing. These are the models that were used in last years NTIRE challenges. See: https://data.vision.ee.ethz.ch/cvl/ntire21/

regarding 1.:
These models are used for upscaling.

regarding 2:
Models are used for denoising and restoration.

Cu Selur
Reply
#3
I performed some tests, for resize BasicVSR++ is too slow to be used in practice.

For resize the best combination is to use NNEDI3 for resize and BasicVSR++ (with model: NTIRE 2021 (3)) to improve the image quality.

In this case the speed is about 5 fps (with RTX3060 and FP16 enabled) and the final quality comparable with RealESERGAN.
Reply
#4
ML stuff is mostly slow, espescially when using higher resolution.

Quote:For resize the best combination is to use NNEDI3 for resize and BasicVSR++ (with model: NTIRE 2021 (3)) to improve the image quality.
Happy you found a meaningful way to use it. Smile

Cu Selur
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)