Does DINOv2 use in inference or only used to guide in training step?

google-research / omniglue

Code release for CVPR'24 submission 'OmniGlue'

https://hwjiang1510.github.io/OmniGlue

Apache License 2.0

532 stars 45 forks source link

Closed thuanaislab closed 3 months ago

thuanaislab commented 3 months ago

Hi, thank you for the great work. Can you answer the question above? Thank you very much.

arjunkarpur commented 3 months ago

Hi, DINOv2 is used during inference. DINOv2 coarse features are used to guide/prune the attention map constructed between the two images' features.