oravus / SeqMatchNet

Code for the CoRL 2021 paper "SeqMatchNet: Contrastive Learning with Sequence Matching for Place Recognition \& Relocalization"
MIT License
56 stars 3 forks source link

Replicability on MSLS dataset #1

Open ga1i13o opened 2 years ago

ga1i13o commented 2 years ago

Hello, I am studying your work together with the SeqNet paper and trying to reproduce results, especially on MSLS. Therefore I would like to kindly ask you to add the .db files for the MSLS dataset, as well as the script to extract the .npy descriptors, just like you thoughtfully did for the SeqNet repo https://github.com/oravus/seqNet Thanks for your help

oravus commented 2 years ago

Hi @ga1i13o,

Thanks for your interest. The process for generating single image NetVLAD descriptors for SeqMatchNet is exactly the same as SeqNet. I will update the Readme in this repo to redirect to SeqNet for that purpose.

Regarding MSLS, we recently included it in SeqNet, so as an immediate remedy, it should be possible to directly borrow and merge Jan 17 commits [a435d36 - 0367a16] using git cherry-pick. If you are happy to wait, I will update this repo next week.