Many people do what you do. You can look at the fuel dataset package that use hdf5. Or you can use three file directly. When you open the file, just the take the slice you want to use. This will return an object you can pass to Theano.
Fred Le 30 sept. 2016 09:24, "Homa Rasouli" <[email protected]> a écrit : > Hey guys, > > I have handered million images on the HDF5 format which I want to feed > them to the network on the fly because it's too big and eats memory. I'm > wondering Theano might not support HDF input. Do you have any other idea > rather than HDF format to store and load images? I also wanna know how can > I split my HDF dataset to chunks for networks? is there any samples that I > can follow to learn ? > > Best, > > -- > > --- > 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. > -- --- 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.
