I agree. The Perceiver is definitely a super interesting paper by Deepmind! Quite promising.

The only thing is that I wanted to talk about transformers applied to computer vision with a model that could be tested right away by the reader, as most of my readers are pretty technical. This is why I chose this one over The Perceiver as they implemented the code and it is easy to use.

Also, my friend Yannic Kilcher covered the Perceiver extremely well on his youtube channel and I did not want to do this again as his explanation is perfect!

The goal was mainly to talk about how transformers CAN be applied to CV, and show an example of how it can be achieved! It will be different in the future and most certainly even different from how Perceiver works.

But thank you for the remark. I completely agree with you.