I'm wondering if they are some state-of-art algorithms to apply the convolutional neural network approach to 3D pictures, eg. input is no longer a grid of pixels, but voxels. My objective is to extract automatically features from 3D pictures and, cherry on the cake, perform supervised learning using 3D pictures as input data. Maybe others approach exist to extract features, but I would like to stay in the scope of machine learning approach (eg. don't want to use methods such as Fourier transform). Thanks in advance for your insights!


1 Answer 1


3D Shape Nets is published this year at CVPR (2015). You could access it at 3D ShapeNets: A Deep Representation for Volumetric Shapes. This is a good way to use voxels as input to DNNs. There are also works applying DNNs to medical imaging such as Deep Neural Networks for Anatomical Brain Segmentation and 3D Deep Learning for Efficient and Robust Landmark Detection in Volumetric Data. I guess these would help you a lot.

  • $\begingroup$ Impressive papers, thanks! Yes, that will help a lot. I'm going to read them carefully. Thank you very much for your help! $\endgroup$
    – Mic
    Commented Aug 5, 2015 at 20:53
  • $\begingroup$ The last link doesn't work. I found the paper using the title I added. $\endgroup$ Commented Jan 22, 2023 at 19:53
  • $\begingroup$ @GeorgeIrwin Best to also edit the link, then. $\endgroup$
    – Peter K.
    Commented Jan 23, 2023 at 3:11
  • $\begingroup$ updated after 7 years ;) $\endgroup$ Commented Mar 2, 2023 at 14:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.