Possible to share Test/Training Data?

[beniroquai]

Hey,

nice work! I wanted to test your network with our data, but was not able to run the test. Do you have the training/test-data somewhere (besides BOX, I don't have an account)? Or could you provide the MAT-structure of the training files?

Thanks a lot and even more for open-sourcing everything! Best Bene

[sunyumark]

Hi Bene

Thank for paying attention to my work. The testing and training data we used are CelebA which can be downloaded here: http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html (aligned and crop). The pre-trained models are unfortunately available via Box. Currently, we are looking for a new place to share our data and I will let you if we find one.

Thanks, Yu

[beniroquai]

Thanks for getting back! The fwd model, namely the lippmann Schwinger model won't be available to process the data, will it?

Best Bene

[sunyumark]

Actually, you need the Lippman Schwinger to generate the measurements. We now are not able to upload the training data to GitHub due to the space limit. We will probably upload them to other places.

^_^

[sunyumark]

Probably the fastest way is to register a box account and download the pre-trained model if you are in a hurry?

[YangFangShu]

Dear Sunyu,

I am also interested in your nice work, and I wanted to test your network, but was not able to do. I also have no BOX account, so that I can not download the training/test-data . I am so appreciated that you would like to share the dataset and the pre-trained model as soon as possible.

Thanks a lot and even more for open-sourcing everything! Best Fangshu

[sunyumark]

I will open-source my pre-trained model via github as soon as possible

[haneneby]

Hi I am also interested in your nice work, and I wanted to test your network but was not able to do. I m wondering how do you create the input images out of the Lippman Schwinger measurements. I mean what is the first backpropagation step? is it pure physics based inverse reconstruction?

Also which tool did you use to compute the competing methods results?

Thanks

Hanene

[beniroquai]

Maybe to add one comment to the thread - the Lippmann-Schwinger in 2D can be found e.g. here: https://github.com/ThanhAnPham/Lippmann-Schwinger/

Am Sa., 9. Feb. 2019 um 17:56 Uhr schrieb haneneby <[email protected]

:

Hi I am also interested in your nice work, and I wanted to test your network but was not able to do. I m wondering how do you create the input images out of the Lippman Schwinger measurements. I mean what is the first backpropagation step? is it pure physics based inverse reconstruction?

Also which tool did you use to compute the competing methods?

Thanks

Hanene

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/sunyumark/ScaDec-deep-learning-diffractive-tomography/issues/1#issuecomment-462086925, or mute the thread https://github.com/notifications/unsubscribe-auth/AEJOuDKttAQDBIcRWZhU1Ldc0LVFrANGks5vL1IEgaJpZM4WZGbq .

[sunyumark]

Hi Hanene,

I've open-sourced the Lippmann-Schwinger as well. You can find the code in the pre-trained models and code

Yu