For pylearn2 runs, need to know what shape the images will be once they've been loaded. Unfortunately, seeing as there's different ways to resize images, will need an invariant way to find out what size the images are expected to be. Think the Settings class in utils should store this information and writing pylearn2 wrapper to use this. Expect this property to be called final_shape and be a 2d tuple (what you would get from a numpy array .shape call).
For pylearn2 runs, need to know what shape the images will be once they've been loaded. Unfortunately, seeing as there's different ways to resize images, will need an invariant way to find out what size the images are expected to be. Think the Settings class in utils should store this information and writing pylearn2 wrapper to use this. Expect this property to be called
final_shape
and be a 2d tuple (what you would get from a numpy array.shape
call).