hoplee6 / rrin-pytorch Goto Github PK

@HopLee6 hi,
In model.py file at line 55, why final output is restricted between 0 and 1? Is there any specific reason? or specific to data range?

Thank You.

@HopLee6
Here is the error message:

Hey, great thanks for this work!

Was checking your source code and found unused variable "final":
https://github.com/HopLee6/RRIN/blob/master/model.py#L55
Is this intended or a bug?

Hey, thank you for your great work.

Do you plan to release code to train model? I am trying to reproduce your results using code from this repo and my training code but it doesn't seem to work.

Hi,

Instead of 2 images, I was wondering what the best way is to have RRIN output interpolated frames for a video. If I convert the video to a sequence of images, should I modify demo.py to loop through the images and pick 2 at a time, or is there a more efficient way to handle videos/image sequences?

Thanks!

您好，想问问您的这个代码输入边长可以不是16的倍数吗 比如宽是180，但我直接送图片进模型他将最后一行像素往下填充了。该怎么改啊

thanks very much~

I tried using this code for Middlebury dataset.
http://vision.middlebury.edu/flow/data/

For most sequences(Beanbags, DogDance, Grove2, Grove3, MiniCooper, Urban2 Urban3, Walking), it worked fine, but for some(Dimetrodon, Hydrangea, RubberWhale, Venus), I got a Runtime error.

RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 1. Got 380 and 368 in dimension 2 at /pytorch/aten/src/THC/generic/THCTensorMath.cu:83

How can I fix this error?

感谢您的工作！我刚刚接触这个领域，很好奇RRIN中的Unet是如何计算光流的？

谢谢您分享的工作，但是我用您提供的代码和模型，在自己的测试代码上做指标对比，UCF101数据集计算的指标如下：Avg PSNR: 32.66783090833283 SSIM: 0.9665849067277201。与您论文的指标PSNR:34.93, SSIM:0.9496不一致。有些疑惑，希望能得到解答。