Hi,

How come you are calculating PSNR between the bicubic image and predicted image? Shouldn't it be the PSNR between ground truth and predicted?

该讨论已关闭

thanks for your great contribution to this project ,but the SR result looks like the same as the original one,even worse,however ,the image size is larger than original one ,am i wrong with run this project?

Did you achieve your posted results using the default parameters only varying the scale, training file and eval file? This would be interesting because I believe the results you posted from the paper were based on the imagenet dataset und not the T91 dataset.

Also, why did you use a custom RGB to Y function? Wouldn't a standard grayscale function do the trick?

Hi,

1. Why did you do zero-padding?
2. Why use Adam instead of the SGD? How come you didn't just use SG like the paper?
3. Why remove the weights initialization? Why remove it?

请问能提供百度网盘或者谷歌网盘吗？Dropbox网盘打不开，谢谢了！
Can you provide baidu net disk or Google net disk? Thanks for failing to open the Dropbox.

Sorry to bother you, but what should I do if I want to calculate the average SSIM?

Hello author，I wonder how much you num-epochs

如果我想根据我自己的低分辨率图像生成高分辨图像，怎样训练呢

请问怎么直接使用预训练的模型 直接进行测试

Hi, Ive ran inference of this code but output image is not resized to x2 or x3.
input image

output image

Is there anything which could be done ? Output image is needed to be resized to 2x..

为什么hr那里就需要插值，不是就下采样后再进行插值作为输入吗，那里训练集的处理不太懂

您好，我是新手
测试的时候需要修改哪些位置，
我已经改了项目说明里的三个路径，
但是argparse返回了一个错误信息，
TypeError: 'required' is an invalid argument for positionals.

with torch.no_grad():
preds = model(y).clamp(0.0, 1.0)

psnr = calc_psnr(y, preds)
print('PSNR: {:.2f}'.format(psnr))

在test中的计算原图与生成的图的psnr也有问题吧？psnr = calc_psnr(y, preds) 这里的y是低分辨率图像的y通道，不是应当计算原图与preds的psnr吗？
有大佬知道吗？

How did you use prepare.py? You say that we can use prepare.py to create custom dataset. But I don't know how to use the output-path.

I found the y channel is multipy 64.738 while I searched the value is 65.738. I wonder what is the right answer. Thanks!

why input channel is 1 rather than 3? I checkout the code and find that the image is converted from rgb to y channel. I don't konw why converted?

Hi，
I had a problem running the eval set.The. H5 file generated by Set14 and BSD100 cannot be read. What is the reason?
and always report errors:
ValueError: Field names only allowed for compound types