search

gmberton / CosPlace

Official code for CVPR 2022 paper "Rethinking Visual Geo-localization for Large-Scale Applications"
MIT License
299 stars 58 forks source link

ViT trained models #33

Closed oravus closed 1 year ago

oravus commented 1 year ago

Hi @gmberton, are ViT backbone models available somewhere to download?

gmberton commented 1 year ago

Hi @oravus, they are not online because with the code only CNN backbones are available (having ViT and CCT was making the code a bit messy). If you want I can pass them to you somehow, but you will need to adjust the cosplace_network.py by yourself because the current version does not support ViT. Let me know, I could upload it temporarily to a GDrive (it's about 330MB).

oravus commented 1 year ago

Thanks, @gmberton. A Gdrive link would be great, I can edit the cosplace_network.py accordingly (guessing vgbench would be helpful for this).

gmberton commented 1 year ago

This is the link to the trained ViT model which I uploaded to GDrive. Let me know when you download it so I'll delete it from there https://drive.google.com/file/d/1yvUhCjCb47ygV6ev3z1g6pJF3Wb_I0K9/view?usp=sharing

oravus commented 1 year ago

Done, many thanks! :)