Volume to scalar classification

Does Pytorch supports volume to scalar classification, i.e, the labels correspond to the 3D input data as a whole, instead of voxel wise classification ?
There is any example?

Thank you!

If I understand you correctly, you would like to classify volumetric data, i.e. 3d input to a class.
You could use Conv3d instead of Conv2d and build a standard conv net like this one.
What kind of data are you using?

1 Like

I am using 3D PET data (medical data). And I have one label (that is one number: class 1, class 2, etc… per volume). This example is also like that ?
Thank you for your replying!

Yes, this example uses 2d images of hand-written digits as input data and classifies them into 10 classes.
It should be a good starter. Let me know, if your model is learning something :wink:

Thank you! Yes, I will let you know :wink: