run on custom dataset

[marius9043]

Hello Dewei,
thank you for sharing your work!
Is it possible to run your code on a custom Dataset? If yes, how would i manage this?

[DeweiHu]

Hi there, Yes, if you have relatively clean data, you can certainly pollute it with customized noise and then try the model to denoise it. In this way, the original image can be leveraged as reference for evaluation.

…

________________________________ From: marius9043 ***@***.***> Sent: Thursday, July 20, 2023 8:58 AM To: DeweiHu/OCT_DDPM ***@***.***> Cc: Subscribed ***@***.***> Subject: [DeweiHu/OCT_DDPM] run on custom dataset (Issue #9) Hello Dewei, thank you for sharing your work! Is it possible to run your code on a custom Dataset? If yes, how would i manage this? — Reply to this email directly, view it on GitHub<#9>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AL6WZQHFBTHUCI2OU2673WLXRE2RTANCNFSM6AAAAAA2RPGMZM>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[aneetachris]

Hi, First of all, I would like to thank the authors for sharing their work. It would have been helpful if you could share some of the images for testing. If we use the Duke dataset, I would be polluting it with speckle noise and performing denoising.
I have seen in the code that the image intensities are converted into the scale of 0 to 3 while training and -1 to 1 during testing. which one should we follow and why? It would be helpful if you could explain regarding this rescaling.

Thankfully,
Aneeta