search

shangqideng / PSRT

22 stars 1 forks source link

About the metrics and loss #1

Open JingYu22 opened 1 year ago

JingYu22 commented 1 year ago

Thanks for sharing, it is a really great job!

I ran the code you provided and the results for CAVE dataset are PSNR:47.3252, SAM:2.6834, SSIM:0.9969, ERGAS:1.8787, I also used the best model you provided and I got PSNR:47.8980, SAM:2.4749, SSIM:0.9977, ERGAS:1.6294, Both are a little worse than the experiment results in paper, that are PSNR:48.78, SAM:2.47, SSIM:0.995, ERGAS:1.66.

Besides the loss used in code is L1 loss, which is different from that in paper. My results are also obtained by L1 loss.

I really hope you can help with my problems.

shangqideng commented 1 year ago

On the one hand, we conduct training under the non-public training framework, which may lead to the difference in training effect, and we just post a simple platform. On the other hand, there may be some differences in data simulation.

JingYu22 commented 1 year ago

I really want to follow your job. Is the non-public version available? I got the data from the download link here, is it different from your data simulation too?

shangqideng commented 1 year ago

You can add my wechat (15881005856) if available, and I will tell you the details about the Data and Training

SeeInTheDark commented 1 year ago

You can add my wechat (15881005856) if available, and I will tell you the details about the Data and Training

You have done a great work, and I also want to follow your job, could I add your wechat?

shangqideng commented 1 year ago

You can add my wechat (15881005856) if available, and I will tell you the details about the Data and Training

You have done a great work, and I also want to follow your job, could I add your wechat? sure, you can contact me via it.

lixin2002cn commented 4 months ago

Thanks for sharing, it is a really great job!

I ran the code you provided and the results for CAVE dataset are PSNR:47.3252, SAM:2.6834, SSIM:0.9969, ERGAS:1.8787, I also used the best model you provided and I got PSNR:47.8980, SAM:2.4749, SSIM:0.9977, ERGAS:1.6294, Both are a little worse than the experiment results in paper, that are PSNR:48.78, SAM:2.47, SSIM:0.995, ERGAS:1.66.

Besides the loss used in code is L1 loss, which is different from that in paper. My results are also obtained by L1 loss.

I really hope you can help with my problems.

Hi, I run the code recently too. I run the result in cave test dataset and only got PSNR 45. Did you meet this state?