I believe instead of using sensors and camera with computer vision, what if someone uses RFID systems for self checkout.

Implementing RFID would be cost effective as compared to implementing Amazon Go based approach. Where you need to install sensors and camera installed!

RFID tags are the major cost, which can be solved with bulk purchases.

Decathlon has already deployed RFID based self checkout system!

Attribuitation is one potential issue/ edge case source that I wonder how they solve. If two people enter/leave the check out area at the same time the "overlap" could get it confused. Machine vision could potentially pair better. Theoretically triangulation could help some but it brings to mind testing it by doing things like having two people walk side by side, pick up a bottle and drape it over th other's shoulders and seeing if it judges them as buying the opposite.

How does that work? Can the checkout machine read the tags inside the bag? I'd prefer if my groceries didn't advertise themselves as I walk down the street.

Checkout machines will have RFID readers. When you place your bag inside RFID reader, all tags will be read.

UHF based RFID readers are better since they have more than 1 meter of read range.

I don't want to build a NFC based readers. Although, I have watched tutorials on NFC reader with Arduino (RC 522 module). It's not that complicated.

My ideal scenario where reader will be implemented is retail stores (Once prototype is working as expected).

Tags would be passive.