Open feifeifei1113 opened 2 years ago
Besides, the file “super_resolution.yaml” needs to be modified according to your own path.
Besides, the file “super_resolution.yaml” needs to be modified according to your own path.
hello, can you save the SR images when Test_model is ASTER? I have some questions. See Issues... my email: foxbeing@hotmail.com
Besides, the file “super_resolution.yaml” needs to be modified according to your own path.
I met some questions when I ran this code.For example,'RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED'occurred after several training epoches.Can you contact me to make a communication,thanks.my email:1487458940@qq.com
I ran it successfully. But the requirements given by the author are incomplete, and you need to install the missing libraries yourself
Why does the eval stage take a particularly long time and half an hour? I trained for 100 epochs, and the highest accuracy is only about 73%. Does anyone know what the problem is?
[2023-05-18 06:34:35] Epoch: [99][1170/2170] vis_dir=TATT/ loss_total: 4.616 loss_im: 4.345 loss_teaching: 0.271 loss_tssi
m: 2.211 0.001
======================================================
evaling easy
[2023-05-18 06:47:42] loss_rec 0.000| loss_im 0.000 PSNR 23.49 | SSIM 0.8663 LPIPS 0.1314
[2023-05-18 06:47:42] PSNR_LR 22.73 | SSIM_LR 0.7977 LPIPS_LR 0.2205
save display images
sr_accuray_iter0: 70.54%
lr_accuray: 62.38%
hr_accuray: 93.39%
AVG inference: 161.2353579121367
sum_images: 1619
best_easy = 72.33%
evaling medium
[2023-05-18 06:59:18] loss_rec 0.000| loss_im 0.000 PSNR 19.21 | SSIM 0.6797 LPIPS 0.2187
[2023-05-18 06:59:18] PSNR_LR 19.24 | SSIM_LR 0.6321 LPIPS_LR 0.3361
save display images
sr_accuray_iter0: 53.15%
lr_accuray: 41.32%
hr_accuray: 86.96%
AVG inference: 161.645155606492
sum_images: 1411
best_medium = 53.15%
evaling hard
[2023-05-18 07:10:03] loss_rec 0.000| loss_im 0.000 PSNR 20.01 | SSIM 0.7408 LPIPS 0.2146
[2023-05-18 07:10:03] PSNR_LR 19.59 | SSIM_LR 0.6663 LPIPS_LR 0.3037
save display images
sr_accuray_iter0: 36.56%
lr_accuray: 31.57%
hr_accuray: 75.65%
AVG inference: 161.92022560144926
sum_images: 1343
best_hard = 37.53%
[2023-05-18 07:10:18] Epoch: [99][1220/2170] vis_dir=TATT/ loss_total: 5.321 loss_im: 4.516 loss_teaching: 0.805 loss_tssi
m: 2.300 0.001
I ran it successfully. But the requirements given by the author are incomplete, and you need to install the missing libraries yourself