Cannot get good result using mobilenet using default settings

[wanghaiboanhui]

No description provided.

[BrainPotter]

I encountered the same problem, and I think the link may not provide the right file.

[wanghaiboanhui]

I have just got 0.67 SSIM and 24 PSNR which is too poor performance comparing to described in paper. Please author provide your trained result best_fpn.h5.

[nbu-lhj]

I can't find the mobilenetv2.pth.tar when use fpn_mobilenet as generator. can you provide it , thanks!

[Williamlizl]

Hi, could you tell me the structure of train sets? error rise,

Traceback (most recent call last): File "train.py", line 183, in <module> train, val = map(get_dataloader, datasets) File "/home/lbc/DeblurGANv2/dataset.py", line 139, in from_config files_a, files_b = map(list, zip(*data)) ValueError: not enough values to unpack (expected 2, got 0)

[oywstju]

Have you figure this problem? I got low SSIM & PSNR on GOPRO, but a very high SSIM & PSNR on my own dataset, I'm so confused.

[daneren]

Hi, could you tell me the structure of train sets? error rise,

Traceback (most recent call last): File "train.py", line 183, in <module> train, val = map(get_dataloader, datasets) File "/home/lbc/DeblurGANv2/dataset.py", line 139, in from_config files_a, files_b = map(list, zip(*data)) ValueError: not enough values to unpack (expected 2, got 0)

How did you solve this problem, I encountered the same problem

[aldenjenkins]

Have you figure this problem? I got low SSIM & PSNR on GOPRO, but a very high SSIM & PSNR on my own dataset, I'm so confused.

I, too, am wondering the same thing

[angelamin]

I can't find the mobilenetv2.pth.tar when use fpn_mobilenet as generator. can you provide it , thanks!

you can just set pretrained=False in fpn_mobilenet.py, since you will load the the train weight later, you don't have to init the mobilenet when testing