How to get VGG-16, Alexnet to handle input sizes of 1000 x 1000?


In going through Week 2 lectures on CNN, 1 thing I didn’t see covered in video lectures was how to get Alexnet or VGG-16 to real life input images which are NOT of size 224 x 224 OR 227 x 227?

e.g. my images are grayscale 1000 x 1000 in dimension? Alexnet takes in 227 x 227 RGB and VGG-16 takes in 224 x 224 RGB. How do I get Alexnet and VGG-16 to handle my inputs?


Resize your images, so they match what the standard model expects and was trained with.

For your grayscale images:
Maybe there is a different standard model you can use for grayscale.
Or you can use your grayscale values for each of the three color planes.