Tianhe Wu
> [本文1.3](https://github.com/memorylorry/PSPNet/blob/asppnet/Train.md)节给出了resnet50_v2.pth下载路径（initmodel目录包含该文件）。 Thanks very much! I also encountered this problem when I was doing pspnet!
Hi, skumra! I have the same question. I don't understand why I turn "ds-shuffle" it will become IW? It just make the index become shuffled.
> Hi, skumra! I have the same question. I don't understand why I turn "ds-shuffle" it will become IW? It just make the index become shuffled. I understand the reason....
Hi! Thanks for you question! **With high resolution images, our methods are not friendly for high resolution images.** If you want to use MANIQA, we advise that you can resize...
Hi! I know your problems! We suggest you to train MANIQA with Kadid10k dataset (because Kadid10k have more kind of images). It can provide more precise image scores.
> 所以你这里输入的是用高斯模糊处理的图像是吗，本应该是插值的是吗 这里作者对输入图像做高斯模糊是没有问题的，论文中作者在Experiments部分有写到为了得到低分辨率图片，作者对其做高斯模糊。
Hi, I know this question. This reason is the problem of implementation. It just support one GPU to training.
One solution is to return the output of each layer directly, without using the SaveOut module
1. This is an error in our paper. Thanks. Each image is test by averaging cropped 20 times results. 2. During the paper writing period, we didn't test our model...
Hi, I just see that. Sorry, I can not get your mean. The split(8:2) is random.