Hi all, Thanks a lot for your feedback! I am a complete newbie to Theano and I first have to get familiar with the organization of the library but, to summarize, you all believe that this is doable in Theano with no modification of the library (the U-net folks mentioned they had to modify Caffe). Whether Lasagne would be simpler (or necessary) layer on top of Theano for this architecture you do not seem to agree.
Do you know any Theano public sample that I could use as a starting to get to something similar to U-net? Also training this type of network on microscopy images will be a big headache and I should most probably consider training set extension strategies such as random affine deformation... is there some utility function for that in Theano or should I consider another Python library, if so which would you recommend for that? For the time being I could configure gpu support under Python 3.4 / Windows... it is already a small step! Best, Sébastien -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
