Closed seainlost81 closed 3 years ago
Hi,
Yes it's confusing, I'll admit. The dataset here is used: https://data.ciirc.cvut.cz/public/projects/2020VisualLocalization/RobotCar-Seasons/
The .txt files provide 6DOF matrix ground truth localizations. I put a query image in (from either the Night or Night-rain image folders), convert to "daytime", find the closest image in the Overcast folder, then compare the localization of the closest Overcast image to that of the Night query.
Thank you very much for your reply, I think I understand what you mean, thank you!
Dear AAnoosheh, thanks for your work. I have a question. I can utilize the NetVLAD to obtain the retrieved image, but I don‘t know how to calculate the threshold accuracy (I know the calculation formula). In other words, I don’t know how to get the location information of the query image and retrieved image. I think they should come from overcast-reference.db and query.db. However, I don't know how to use them. Thanks very much if you share that. My email address is liuhaibo81@gmail.com, thanks in advance.