opteroncx

To solve this bug, simply edit line147-149 1. remove code like 'np.uint8()' 2. convert to bgr and use opencv for output

Use div2k, Random rotate or flip the input matrix in dataloader (augmentation) train with 32x32 input for 200 epochs, set lr=1e-4 and decrease 10 times after 100 epochs. then use...

for example: def data_augment(im,num): org_image = im.transpose(1,2,0) if num ==0: ud_image = np.flipud(org_image) tranform = ud_image elif num ==1: lr_image = np.fliplr(org_image) tranform = lr_image elif num ==2: lr_image =...

Decrease the learning rate to 1e-5, and step set to 10

比赛的时候是从小到大（分支由少到多）一级一级finetuning的，最后再一起加上nonlocal，大概3个分支开始，每次50个epoch。AIM这个数据时华为伦敦研究院做的合成数据，如果用真实数据（比如TIP18那个）模型啥也不改直接切中间区域跑大概28.1左右。如果需要图形化的测试环境，可以看看我另一个坑https://github.com/opteroncx/MoePhoto

如果对二次元和图像修复感兴趣欢迎加群320785467讨论

Hi, you could generate the .h5 file with these matlab files or you could rewrite your dataloader.py to fit your own data. https://github.com/twtygqyy/pytorch-LapSRN/tree/master/data You could also use the --pretrained method...

The result of PSNR and SSIM in the tabel was computed from the whole image, the PSNR and SSIM in the figure of visual comparison was computed from each sub...

Please show me the full error log and I will check if there are some bugs in the code.

imglist[i] ---> imglist[-1]