Dachun Kai

BasicVSR原文实验部分和EDVR在vimeo-test中对比时提到，“This is expected since Vimeo-90K-T contains sequences with only seven frames, while the success of BasicVSR partially comes from the aggregation of long term information (which is a realistic assumption).”...

此处[代码](https://github.com/XPixelGroup/BasicSR/blob/master/basicsr/models/video_base_model.py#L133-L139)中，默认假设了不同folder的len是相同的，仅适用于reds。然而实际验证时并不一定。

How to visualize aligned features like Figure 8 in the paper? Can provide related code?

1

Like this: [Fig 8](https://i.imgtg.com/2023/02/08/cT31U.png). Is it possible to provide related code?

What is GT when training RealBasicVSR

1

Because RealBasicVSR is evaluate by non-reference metric (NIQE, PI), what is the GT or loss used for training RealBasicVSR?

怎么预训练Liteflownet的？

1

请问在vimeo上预训练Liteflownet，是用I1 warp 到 I2，再和I2的真实值算loss吗？这部分无监督的预训练代码可以提供吗？谢谢

Other flow teacher network?

3

May I ask if you have tried using RAFT or other methods as the teacher network? Would the results be better? How did you pretrain LiteFlowNet on Vimeo90k, and could...

Concerns about comparisons

1

Excellent work! Would it be appropriate to consider comparing it with two-stage methods that incorporate enhancement algorithms for various common corruption types along with depth estimation? For example, when dealing...