Preprocessing data

[linhlpv]

Hi,
Thank you so much about the great works!

As I know in many GAN based methods, normalize input to range -1 1 is a tip for stable training but I do not see norm func in your CT-dataset. Could you explain that why you did not use this, plz?

Thank you a lot and hope to hear from you soon.
Best,

[ristea]

Hi! We did normalization based on prior about data. In data/ct_dataset.py you can observe it at lines 31-32, 39-40.

[linhlpv]

hello.
Thank for your reply. So is 1e3 global statistic param of your dataset? I want to know about its meaning.
Thank u.

[ristea]

The raw data are DICOM files in Hounsfield scale. The values for CTs are approximatively in [0, 3500]. This is the reason for that value.