0

I am very unfamiliar with Caffe. My task is to train an autoencoder net on image pairs, given in .tif format, where one is a grayscale image of nerves, and the other is the corresponding binary mask which shows if a certain structure is present on the image or not. I have these in the same "train" folder. What I would like to accomplish, is a meaningful experiment with these images (segmentation, classification, it is not specified). My first problem is that I do not know how to feed the images into the net without an existing train.txt. Can I use the images directly, or another format like lmdb, hdf5 needed? Any suggestion is appreciated.

Pásztor András
  • 3
  • 3
  • are you sure you want an "autoencoder"? if you have labels, it seems like you are after a classification conv net – Shai Sep 04 '16 at 10:10
  • Thank you for the reply. This was not my decision, the goal is here to become familiar with architecture and data handling. So even if it is not the best option, I have to do this now. I did find autoencoder solution for MNIST classification, so I suspect that it is possible. My problem is that I do not have a .txt for lmdb creation just the images.. – Pásztor András Sep 04 '16 at 14:55
  • This answers your question on how to do this in Python: http://stackoverflow.com/a/39097123/5465000 – Amir Sep 05 '16 at 20:14

1 Answers1

0

you can accomplish it with simple classification (existing like alexnet, googlenet, lenet). You can use only the binary mask or gray scale image and the class name to do this. Nvidia Digits is a good graphical tool to make the pair dataset and learning....

Please see this link:

https://developer.nvidia.com/digits

MD. Nazmul Kibria
  • 1,080
  • 10
  • 21
  • I really appreciate your answer. My problem is that I can only use Caffe in CPU mode - as I understand DIGITS requires a different Caffe build? – Pásztor András Sep 04 '16 at 15:28
  • DIGITS do not allow to run caffe on CPU mode, so you need to build/install caffe from source code. Install caffe if you do not build it: http://caffe.berkeleyvision.org/installation.html . – MD. Nazmul Kibria Sep 06 '16 at 04:36