serizba / cliquemining

Close, But Not There: Boosting Geographic Distance Sensitivity in Visual Place Recognition
GNU General Public License v3.0
28 stars 3 forks source link

Model's weights #4

Open gmberton opened 4 months ago

gmberton commented 4 months ago

Hey @serizba! Are you planning to release the weights before ECCV? I would like to try this model on our MeshVPR pipeline, right now SALAD is the model that gives the best results but I think this would work even better

serizba commented 4 months ago

Hi @gmberton,

Yes, I am planning to do so. I will upload them in a couple of days, sorry for the delay.

serizba commented 3 months ago

Hi @gmberton

I've just posted in the readme a link to the weights. Let me know if something is missing or you encounter any issue!

gmberton commented 3 months ago

Thank you Sergio, that was fast! Is there also a way to load the models through torch.hub? I see the hubconf.py file but I think it's the same as SALAD