search

AnyLoc / Revisit-Anything

Code release for Revisit Anything: Visual Place Recognition via Image Segment Retrieval (ECCV 2024)
https://revisit-anything.github.io/
BSD 3-Clause "New" or "Revised" License
108 stars 5 forks source link

Reproduce the results #4

Closed kaigelee closed 1 month ago

kaigelee commented 1 month ago

I would like to ask what hardware environment is needed to reproduce all the results of the paper? For example, 1*3090?

Shubodh commented 1 month ago

Hello @kaigelee , 3090 should be more than sufficient to reproduce all results of the paper. We extract SAM masks at a lower resolution (without much loss in performance) so even 12 GB vRAM should be sufficient (possibly even 8). DINOv2 should run easily at higher resolution too.

Closing this issue. Please feel free to reopen if this doesn't solve your issue.