MPRNet
MPRNet copied to clipboard
swz30
[CVPR 2021] Multi-Stage Progressive Image Restoration. SOTA results for Image deblurring, deraining, and denoising.
在测试MPRNet和先前的deblurGanV2,我使用的是有人像的图片.发现MPRNet有磨皮现象,这样会改变图片应有的信息(会影响后续其他处理判断),而deburGanV2并没有磨皮现象, 
hello, I degraded the image in FFHQ and want to use the debur process in MPRnet to restore it. But, when I train the model, I first met this problem...
Will the different versions of matlab you use for testing affect the results? What is the version number of matlab you are using?
Why is the memory size of the model I trained is 41MB, while the pre-trained model provided is only 13MB?
Where can I test the PSNR and SSIM of my own trained model? The test.py file just generated the recovered image, not testing PSNR and SSIM
About training, i have put the dataset into the required structure directory, why this problem happens several times?? Can anyone tell me how to do??
Hi! Thank you for your wonderful work and great code! Have you tried mixed precision training? I use mixed precision when I train to deblurring model on gopro datasets. But...
good job！ I am training the denoising task, the training environment is the same as the author's, but the size of the trained model is 180M instead of the 60M...
训练模型的问题
 您好，在去雾网络中，我按照您默认的通道数进行训练，可以我训练得到的模型的大小为240M左右，但您提供的预训练模型的大小为80M左右，你能告诉是为什么吗？或许您在保存模型的时候使用了什么样的策略来压缩模型吗？
