Amar Ali-bey

Hello @libenchong, Thank you again for your valuable feedback. Yes, I've noticed some errors in that regard but never did a thorough investigation like you did. That would make a...

Hello @libenchong, Great observation! In most benchmarks, the ground truth (GT) typically includes images located within a 25-meter perimeter from the query, based on GPS coordinates. However, since Mapillary images...

Hello @phy12321, We only trained on images resized to 320x320. We don't provide weights for higher resolution images (even though, MixVPR achieves highers recalls with images resized to 400x400 or...

Hello @wpumain, We have made available the code for testing on 8 different benchmarks (Tokyo 247 not included). To evaluate methods, you can utilize the provided pretrained weights and code...

Hello @Anuradha-Uggi, We have discussed the loss function and online mining strategies in another paper, which can be found at (https://github.com/amaralibey/gsv-cities). The motivation behind collecting the GSV-Cities dataset was the...

It is challenging to establish a definitive way for determining the positive images for each query in Pitts30k. While the potential positive images are those within a 10-meter radius, there...

@Anuradha-Uggi, Relying solely on UTMs is not sufficient as they do not account for bearing (or orientation with respect to the North Pole). For instance, if two images are 5...

Hello @yicocc, Can you share more information about this issue? - Did you download the Nordland dataset as explained? - Did you properly load the pre-trained weights? - Did you...