Shape of an image input in resnet

(Francisco José Tornay Mejías) #1

What shape does a resnet convnet expect input images to have?
I’m using HxWxD (height times width x depth, depth being the three color channels), which is the way numpy encodes jpeg images. However the code in tye pytorch transfer learning tutorial ( seems to expect depth to be the first axis.


In keras I use this HxWxD