A CNN Model that creates filters bu unsupervised learning. The model tries to distinguish the output of each convolution operation and tries to create filters that generate most various output layers.
The theory is to generate convolution layers which are not similar to each other. By doing so, it is aimed to create filters that catch most valuable patterns.
First video: https://www.youtube.com/watch?v=axqSGXU5qIw
Second video: https://www.youtube.com/watch?v=HxO4ncpN7ZA&t=148s
Demonstration 1: https://www.youtube.com/watch?v=zIReaxDhfcQ