Submission details
Method name | NetVLAD (top-50) & D2-Net - multi-scale |
Dataset | silda |
Submission date | Sept. 26, 2019, 7:05 a.m. |
Last modified | Nov. 1, 2019, 9:50 a.m. |
Publication url | http://openaccess.thecvf.com/content_CVPR_2019/papers/Dusmanu_D2-Net_A_Trainable_CNN_for_Joint_Description_and_Detection_of_CVPR_2019_paper.pdf |
Code url | https://github.com/mihaidusmanu/d2-net |
Other info | Uses NetVLAD (bottom part cropped - 825x1000) to identify the top-50 closest database images, then uses multi-scale D2-Net features (bottom part cropped - 825x1000) for feature matching and COLMAP's image_registrator for localization. |
Detailed results for all conditions
Method | evening | snow | night |
---|---|---|---|
NetVLAD (top-50) & D2-Net - multi-scale | 29.6 / 67.8 / 94.8 | 6.0 / 16.4 / 72.3 | 25.6 / 51.6 / 79.9 |