Submission ID: 00020
D2-Net (multi-scale), DEGENSAC
Processed: 20-04-23. Download link: sid-00020-d2net-multiscale-2k.json
This page ranks the submission against all others using the same number of keypoints, regardless of descriptor size. Please hover over table headers for descriptions on metrics and full scene names.
Metadata
- Authors: Challenge organizers (contact)
- Keypoint: d2net-multiscale
- Descriptor: d2net-multiscale (512 float32: 2048 bytes)
- Number of features: 2048
- Summary: D2-Net, multi-scale model, up to 2048 features. Trained on the MegaDepth dataset, removing scenes which overlap with the Phototourism test set. Stereo with DEGENSAC and optimal parameters. Parameters tuned separately from the 8k submission.
- Paper: http://openaccess.thecvf.com/content_CVPR_2019/papers/Dusmanu_D2-Net_A_Trainable_CNN_for_Joint_Description_and_Detection_of_CVPR_2019_paper.pdf
- Website: https://github.com/mihaidusmanu/d2-net
- Origin: Baseline
- Flags: is_baseline
Phototourism / Stereo track
mAA at 10 degrees: 0.12270 (±0.00048 over 3 run(s) / ±0.07372 over 9 scenes)
Rank (per category): 98 (of 108)
| Scene | Features | Matches (matcher) |
Matches (filter) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| bm | 2048.0 | 247.8 | 247.8 | 110.9 | 0.179 Rank: 101/108 |
0.370 Rank: 101/108 |
0.01759 (±0.00036) Rank: 98/108 |
0.04529 (±0.00025) Rank: 99/108 |
| fcs | 2048.0 | 320.9 | 320.9 | 131.9 | 0.148 Rank: 101/108 |
0.315 Rank: 101/108 |
0.09927 (±0.00139) Rank: 98/108 |
0.20442 (±0.00198) Rank: 98/108 |
| lms | 2033.3 | 332.7 | 332.7 | 152.0 | 0.224 Rank: 100/108 |
0.298 Rank: 100/108 |
0.16916 (±0.00264) Rank: 96/108 |
0.28453 (±0.00153) Rank: 96/108 |
| lb | 2048.0 | 288.7 | 288.7 | 116.7 | 0.177 Rank: 101/108 |
0.268 Rank: 101/108 |
0.06938 (±0.00036) Rank: 98/108 |
0.13600 (±0.00240) Rank: 98/108 |
| mc | 2048.0 | 207.0 | 207.0 | 80.0 | 0.151 Rank: 101/108 |
0.339 Rank: 101/108 |
0.02804 (±0.00036) Rank: 98/108 |
0.06178 (±0.00102) Rank: 100/108 |
| mr | 2048.0 | 280.1 | 280.1 | 127.7 | 0.183 Rank: 100/108 |
0.399 Rank: 101/108 |
0.03058 (±0.00072) Rank: 99/108 |
0.06773 (±0.00200) Rank: 100/108 |
| psm | 2044.4 | 243.5 | 243.5 | 118.6 | 0.124 Rank: 101/108 |
0.106 Rank: 101/108 |
0.05031 (±0.00117) Rank: 100/108 |
0.12724 (±0.00088) Rank: 95/108 |
| sf | 2045.8 | 290.1 | 290.1 | 113.6 | 0.144 Rank: 101/108 |
0.249 Rank: 101/108 |
0.02996 (±0.00057) Rank: 100/108 |
0.06861 (±0.00023) Rank: 100/108 |
| spc | 2046.5 | 273.8 | 273.8 | 114.2 | 0.180 Rank: 100/108 |
0.296 Rank: 101/108 |
0.05068 (±0.00193) Rank: 98/108 |
0.10866 (±0.00324) Rank: 98/108 |
| avg | 2045.6 | 276.1 | 276.1 | 118.4 | 0.168 Rank: 101/108 |
0.293 Rank: 101/108 |
0.06055 (±0.00057) Rank: 98/108 |
0.12270 (±0.00048) Rank: 98/108 |
We show the inliers that survive the robust estimation loop (i.e. RANSAC), or those supplied with the submission if using custom matches, and use the depth estimates to determine whether they are correct. We draw matches above a 5-pixel error threshold in red, and those below are color-coded by their error, from 0 (green) to 5 pixels (yellow). Matches for which we do not have depth estimates are drawn in blue. Please note that the depth maps are estimates and may contain errors.
— british museum —
— florence cathedral side —
— lincoln memorial statue —
— london bridge —
— milan cathedral —
— mount rushmore —
— piazza san marco —
— sagrada familia —
— saint paul's cathedral —
Phototourism / Multiview track
mAA at 10 degrees: 0.28297 (±0.00254 over 3 run(s) / ±0.18284 over 9 scenes)
Rank (per category): 99 (of 108)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| bm | 2048.0 | 247.07 | 96.68 Rank: 98/108 |
1540.69 Rank: 37/108 |
2.969 Rank: 102/108 |
1.18831 Rank: 98/108 |
0.01647 (±0.00130) Rank: 100/108 |
0.04751 (±0.00144) Rank: 100/108 |
| fcs | 2048.0 | 330.83 | 95.63 Rank: 77/108 |
2393.50 Rank: 17/108 |
3.024 Rank: 101/108 |
0.49176 Rank: 97/108 |
0.27570 (±0.00833) Rank: 99/108 |
0.42004 (±0.00959) Rank: 98/108 |
| lms | 2033.3 | 358.10 | 95.65 Rank: 92/108 |
2688.60 Rank: 2/108 |
3.314 Rank: 99/108 |
0.38953 Rank: 94/108 |
0.48693 (±0.00989) Rank: 97/108 |
0.62907 (±0.01384) Rank: 97/108 |
| lb | 2048.0 | 340.45 | 97.12 Rank: 62/108 |
2263.71 Rank: 7/108 |
3.243 Rank: 101/108 |
0.61952 Rank: 101/108 |
0.19892 (±0.00683) Rank: 98/108 |
0.34362 (±0.01253) Rank: 97/108 |
| mc | 2048.0 | 206.04 | 90.96 Rank: 102/108 |
1378.05 Rank: 41/108 |
2.924 Rank: 101/108 |
0.84862 Rank: 101/108 |
0.09817 (±0.00081) Rank: 100/108 |
0.19337 (±0.00323) Rank: 100/108 |
| mr | 2048.0 | 281.27 | 87.13 Rank: 89/108 |
1619.45 Rank: 47/108 |
3.213 Rank: 99/108 |
0.93364 Rank: 100/108 |
0.06396 (±0.00689) Rank: 99/108 |
0.12517 (±0.00987) Rank: 100/108 |
| psm | 2044.4 | 240.08 | 77.29 Rank: 65/108 |
1976.64 Rank: 28/108 |
2.256 Rank: 98/108 |
0.94807 Rank: 99/108 |
0.01937 (±0.00090) Rank: 96/108 |
0.06071 (±0.00230) Rank: 95/108 |
| sf | 2045.8 | 288.44 | 92.48 Rank: 93/108 |
2131.35 Rank: 26/108 |
2.992 Rank: 101/108 |
0.57943 Rank: 101/108 |
0.17957 (±0.01378) Rank: 100/108 |
0.30102 (±0.01658) Rank: 100/108 |
| spc | 2046.5 | 282.64 | 96.91 Rank: 80/108 |
2002.33 Rank: 23/108 |
3.179 Rank: 100/108 |
0.96937 Rank: 99/108 |
0.26016 (±0.00189) Rank: 100/108 |
0.42620 (±0.00465) Rank: 98/108 |
| avg | 2045.6 | 286.10 | 92.21 Rank: 86/108 |
1999.37 Rank: 19/108 |
3.013 Rank: 101/108 |
0.77425 Rank: 99/108 |
0.17770 (±0.00162) Rank: 99/108 |
0.28297 (±0.00254) Rank: 99/108 |
In the multi-view track we reconstruct the scene with Structure-from-Motion (Colmap) with small sets of images. We show the results for one bag of 25 images (displaying: 10). Keypoints are drawn in blue if they are part of the model, and in red otherwise.
— british museum —
— florence cathedral side —
— lincoln memorial statue —
— london bridge —
— milan cathedral —
— mount rushmore —
— piazza san marco —
— sagrada familia —
— saint paul's cathedral —
