search

gmberton / deep-visual-geo-localization-benchmark

Official code for CVPR 2022 (Oral) paper "Deep Visual Geo-localization Benchmark"
MIT License
187 stars 28 forks source link

reproduce the result of CCT-NetVLAD #2

Closed Lu-Feng closed 2 years ago

Lu-Feng commented 2 years ago

Hi, I really appreciate for the work you've done. I wonder if the CCT-NetVLAD model will be released. How should I set parameters (trunc_te/freeze_te) to reproduce the result of CCT-NetVLAD on MSLS dataset.

rm-wu commented 2 years ago

Hi @Lu-Feng, Thanks for your interest in our work! To reproduce the results with CCT+NetVLAD on MSLS, you must set --trunc_te=8 --freeze_te 1. We will update the trained model soon! Thanks for pointing out that it is missing!

Lu-Feng commented 2 years ago

Hi @Lu-Feng, Thanks for your interest in our work! To reproduce the results with CCT+NetVLAD on MSLS, you must set --trunc_te=8 --freeze_te 1. We will update the trained model soon! Thanks for pointing out that it is missing!

Thanks for your reply! I have reproduced the result with your help.

ga1i13o commented 2 years ago

Hello, I uploaded the pre-trained models, with a link in the README. An additional note about the results with CCT, you can notice an improvement specifying a "--negs_num_per_query" lower than 10, like 2 or 5