If we could find out the selection criteria behind each layer of the neural network for the human visual cortex we could possibly build something more accurate.

If we could find out the selection criteria behind each layer of the neural network for the human visual cortex we could possibly build something more accurate. Although I doubt the visual cortex is a simple feed forward network like the one used in the paper. It's likely to have a non linear structure that's significantly more complex.