I have a bunch of images taken from a camera showing a pipe and would like to detect if the pipe is leaking or not. There are very few examples of leaking pipes in the data set. So considering this problem as a supervised learning problem, I think that it may not give us good results due to imbalanced data. I am thinking of using autoencoders and considering it as an outlier detection problem.
I am new to deep learning so I would like to know what the architecture of my neural network should look like. Should I have some convolutional layers first and then an autoencoder or should I only have an autoencoder? What would be the best deep learning library for such a use case? I am also thinking of using only the photos which do not have any leak for the training phase, is that okay?