增强后的视频存在白边

[cslvjt]

感谢作者对视频增强领域提出的贡献，非常有意义！对我帮助很大！我又下面的问题想请教一下：
现象：我这边跑你的模型时发现增强后的视频存在白边，我发现你们给的视频中也存在这种情况https://youtu.be/hx0zrql-SrU?t=38，
我怀疑是代码中pad的方式为constant=1导致的，但是改为其他pad 模式后，例如reflect，模型增强效果就会发生改变。
针对上面现象请问你们有发现嘛？是否有一个合理的结论？

Thank you for your valuable contributions to the field of video enhancement — your work has been incredibly helpful to me!

I have encountered an issue when running your model: I noticed that the enhanced videos often have white borders around the edges. Interestingly, I found the same phenomenon in the sample video you provided (for example, at (https://youtu.be/hx0zrql-SrU?t=38)). I suspect this might be related to the way padding is handled in your code, where pad is set as constant=1. However, when I tried changing the padding mode to something else, such as reflect, the enhancement results differed significantly.

I would like to ask: Have you also observed this issue? Is there a reasonable explanation or recommended solution for this phenomenon?

Thank you very much, and I look forward to your insights!

[ChenDoubleJ]

感谢作者对视频增强领域提出的贡献，非常有意义！对我帮助很大！我又下面的问题想请教一下： 现象：我这边跑你的模型时发现增强后的视频存在白边，我发现你们给的视频中也存在这种情况https://youtu.be/hx0zrql-SrU?t=38， 我怀疑是代码中pad的方式为constant=1导致的，但是改为其他pad 模式后，例如reflect，模型增强效果就会发生改变。 针对上面现象请问你们有发现嘛？是否有一个合理的结论？

Thank you for your valuable contributions to the field of video enhancement — your work has been incredibly helpful to me!

I have encountered an issue when running your model: I noticed that the enhanced videos often have white borders around the edges. Interestingly, I found the same phenomenon in the sample video you provided (for example, at (https://youtu.be/hx0zrql-SrU?t=38)). I suspect this might be related to the way padding is handled in your code, where pad is set as constant=1. However, when I tried changing the padding mode to something else, such as reflect, the enhancement results differed significantly.

I would like to ask: Have you also observed this issue? Is there a reasonable explanation or recommended solution for this phenomenon?

Thank you very much, and I look forward to your insights!

我进行了测试，当改成reflect时，生成的结果会发生局部的扭曲，请问您会出现这个问题吗，有什么方法能够解决呢？

[cslvjt]

感谢作者对视频增强领域提出的贡献，非常有意义！对我帮助很大！我又下面的问题想请教一下： 现象：我这边跑你的模型时发现增强后的视频存在白边，我发现你们给的视频中也存在这种情况https://youtu.be/hx0zrql-SrU?t=38， 我怀疑是代码中pad的方式为constant=1导致的，但是改为其他pad 模式后，例如reflect，模型增强效果就会发生改变。 针对上面现象请问你们有发现嘛？是否有一个合理的结论？
Thank you for your valuable contributions to the field of video enhancement — your work has been incredibly helpful to me!
I have encountered an issue when running your model: I noticed that the enhanced videos often have white borders around the edges. Interestingly, I found the same phenomenon in the sample video you provided (for example, at (https://youtu.be/hx0zrql-SrU?t=38)). I suspect this might be related to the way padding is handled in your code, where pad is set as constant=1. However, when I tried changing the padding mode to something else, such as reflect, the enhancement results differed significantly.
I would like to ask: Have you also observed this issue? Is there a reasonable explanation or recommended solution for this phenomenon?
Thank you very much, and I look forward to your insights!

我进行了测试，当改成reflect时，生成的结果会发生局部的扭曲，请问您会出现这个问题吗，有什么方法能够解决呢？

我们这边没有想到一个合理的解决方案