Recognize Anything: A Strong Image Tagging Model

recognize-anything.github.io

19 pointsMitPitt4 comments

It looks like the comparison images at the top have inaccurate captions?

RAM looks correct to me. What's wrong?

E.g BLIP image is captioned "missing: dog" despite the dog being detected in the image.

That's why it's a comparison. It's a page about RAM and how it's better than BLIP