Hello, I am aware you merged your code to the OpenMMLab framework. Is there a plan to also merge it to BasicSR?
Thanks.

Hi,

I see basicVSR++ win the NITRE2021 super reslution track, whose visual quality is amazing.
While basicVSR already open source in MMEditing, would basicVSR++ be open sourced soon?

Hello kelvin,

Thanks for your great work! I am very interest in your work, BasicVSR and BasicVSR++, and I have some questions.

In paper, you provided comprehensive comparison of existing VideoSR methods. I noticed that some methods, such as RBPN and EDVR did not have BD experiments in original paper, and methods such as DUF and PFNL did not have BI experiments in their paper.

As for the results of these methods, did you retrain them from scratch?

Best regards,
Jingbo

Hello,

Thanks for making your code public, wonderful work.

I've read that you & your team further improved the architecture as shown here. Are there any plans of releasing basicVSR++ models in the future?

Best Regards,
Jeremi Podlasek

I am currently a second-year Ph.D. student in Harbin Institute of Technology. I am supervised by Prof. Zuo Wangmeng. I am greatly interested in the work of GLEAN, could you share the code? Looking forward to your reply as soon as possible!

Thanks for you work!
Could you please release the pretrained model to mmedit model zoo?

I am interested in your work, and wish you to release the code soon.

Hi,
Thanks for your excellent work.
I'm curious about some details about your evaluation settings on vimeo-90k-T. Are the PSNR and SSIM calculated on all the frames, or just on the 4th frame? Did you use the flip strategy to produce 14-frame sequences during evaluation?
Thanks again.

There are lots of methods about the optical flow. Most of them used Pyramid Network. The shape(size) of pictures in REDS is (180,320). How to use those networks to valid the pictures in REDS (180 % 8 != 0) ? If the downsample factor is 4, the problem can be solved, but I cannot use the pre-trained models which are provided officially in this way.
So, how do you solve it with SpyNet ? I am curious about it.
Thanks!

Hello，what an amazing work! I am very interested in this work. I hope you can release the train codes and the results soon! Thank you very much!

Hi, thank you very much for your excellent work. I have some question about the training deails:

1. As we know, the input of the network is [B, T, C, H, W], does T remain the same during training and inference? If T=100 during training, it's really slow.
2. Detail about the cosine annealing scheme: could you please tell me about the T_period? Did you use the restarts method?

Thank you again for your nice work.

Welcome update to OpenMMLab 2.0

I am Vansin, the technical operator of OpenMMLab. In September of last year, we announced the release of OpenMMLab 2.0 at the World Artificial Intelligence Conference in Shanghai. We invite you to upgrade your algorithm library to OpenMMLab 2.0 using MMEngine, which can be used for both research and commercial purposes. If you have any questions, please feel free to join us on the OpenMMLab Discord at https://discord.gg/amFNsyUBvm or add me on WeChat (van-sin) and I will invite you to the OpenMMLab WeChat group.

Here are the OpenMMLab 2.0 repos branches:

Attention: please create a new virtual environment for OpenMMLab 2.0.

In paper, "To demonstrate the effects, we compare BasicVSR (using bidirectional propagation) with its unidirectional
variant (with comparable network complexity)."please, is the complexity of the network structure of the comparison experiment different here? for example, bidirectional propagation contains 30 resblocks ,unidirectional propagation contains 60 resblocks ? What is the network structure of this comparison experiment（unidirectional propagation）?

Hi, it's a really excellent work. When will you publish the code? Will it be published In 2020? Thanks a lot!

Hi,

How can the model be used to an arbitrary video?

Hi Kelvin,

Thanks for your work and open source effort.

May I ask how long does it take for you to train BasicVSR on the REDS dataset? On my end, the eta is 4 days when training it on 2 Quadro RTX 8000. Is this a normal behaviour?