This code helps us understand how spatial pooling pyramid (SPP) works in context of AlexNet architecture. The input image size is 227 x 227 as required by the architecture. Using the optional SPP, you can experiment with different image sizes and verify that irrespective of the image size(although for an architecture this deep we should have images atleast of size 163 x 163 ),say varying image sizes in a batch, we still should get the same fixed length vector at the output of SPP. You might be able to get results for even images with smaller dimensions if you make the network shallower.Run alexnet_feedforward.py to achieve this