Getting error because of training images are different from testing images ... Please Help

[Shrabani Ghosh]

In my caffe program, I have trained my neural network with 3D brain skull  axial ct images. Now for testing purpose I am using  3D axial, sagittal and coronal skull ct images.to get high resolution images output using  super resolution algorithm. I need to know will it be any problem to get output because I am using axial images  for training purpose and axial, sagittal and coronal  images for testing?

Thanks in advance

[Jonathan R. Williford]

Usually neural networks don't generalize to new types of data. If you want to test on sagittal and coronal images, you should also train on them. This might be less of an issue with super resolution algorithms, but I would still sample the training and testing from the same distribution. Since the images are 3D, I'm not sure what the difference between axial, sagittal, and coronol means. Are the images just rotated? You can apply rotation to the training data to augment your data.

Best,

Jonathan

On Thu, Sep 21, 2017 at 7:15 AM, Shrabani Ghosh <srbn.g...@gmail.com> wrote:

In my caffe program, I have trained my neural network with 3D brain skull  axial ct images. Now for testing purpose I am using  3D axial, sagittal and coronal skull ct images.to get high resolution images output using  super resolution algorithm. I need to know will it be any problem to get output because I am using axial images  for training purpose and axial, sagittal and coronal  images for testing?

Thanks in advance

--
You received this message because you are subscribed to the Google Groups "Caffe Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to caffe-users+unsubscribe@googlegroups.com.
To post to this group, send email to caffe...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/caffe-users/201ae75f-9e88-4a9f-bc5b-916a2c68be30%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.