Which image sizes allow for consecutive  convolution layers with pooling?

Working with labeled arrows

Data sample:

Example where first arrow is up:

Searching for concepts in embedding

Example of moving in concept space

Finding a concept