Submission details

Method nameDenseVLAD (top-50) & D2-Net - multi-scale
Datasetaachen
Submission dateSept. 20, 2019, 5:08 p.m.
Last modifiedNov. 1, 2019, 9:49 a.m.
Publication urlhttp://openaccess.thecvf.com/content_CVPR_2019/papers/Dusmanu_D2-Net_A_Trainable_CNN_for_Joint_Description_and_Detection_of_CVPR_2019_paper.pdf
Code urlhttps://github.com/mihaidusmanu/d2-net
Other infoUses DenseVLAD (single-scale) to identify the top-50 closest database images, then uses multi-scale D2-Net features for feature matching and COLMAP's image_registrator for localization.

Detailed results for all conditions

Methoddaynight
DenseVLAD (top-50) & D2-Net - multi-scale83.1 / 90.9 / 95.540.8 / 62.2 / 80.6