@semordnilap have you tried it with a smaller NN?
1000 training samples seems a bit too low for this network size. In particular the dense layer. But this is just a guess, as I'm not that experienced with deep learning

@Halbeard In the end I solved by using the flow_from_directory function that takes care of preprocessing the images, although I'm still unsure about what I was doing wrong with the labels or training set

Sign in to participate in the conversation
LinuxRocks.Online

Linux Geeks doing what Linux Geeks do..