Submission ID: 00034
D2-Net (single-scale), DEGENSAC
Processed: 20-04-22. Download link: sid-00034-d2net-singlescale-8k.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-singlescale
- Descriptor: d2net-singlescale (512 float32: 2048 bytes)
- Number of features: 8000
- Summary: D2-Net, single-scale model, up to 8000 features. Trained on the MegaDepth dataset, removing scenes which overlap with the Phototourism test set. Stereo with DEGENSAC and optimal parameters
- 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: 0.18499 (±0.00022 over 3 run(s) / ±0.08992 over 9 scenes)
Rank (per category): 136 (of 147)
| Scene | Features | Matches (matcher) |
Matches (filter) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| bm | 4887.9 | 624.9 | 624.9 | 220.7 | 0.295 Rank: 147/147 |
0.442 Rank: 143/147 |
0.03267 (±0.00165) Rank: 133/147 |
0.06317 (±0.00118) Rank: 133/147 |
| fcs | 5450.1 | 787.8 | 787.8 | 281.1 | 0.263 Rank: 147/147 |
0.415 Rank: 142/147 |
0.20162 (±0.00228) Rank: 137/147 |
0.33100 (±0.00288) Rank: 136/147 |
| lms | 4539.5 | 695.8 | 695.8 | 237.7 | 0.281 Rank: 142/147 |
0.331 Rank: 142/147 |
0.21775 (±0.00164) Rank: 134/147 |
0.33353 (±0.00095) Rank: 134/147 |
| lb | 5132.1 | 639.1 | 639.1 | 208.6 | 0.293 Rank: 142/147 |
0.341 Rank: 142/147 |
0.12398 (±0.00165) Rank: 133/147 |
0.20422 (±0.00370) Rank: 134/147 |
| mc | 5960.5 | 722.8 | 722.8 | 257.9 | 0.326 Rank: 147/147 |
0.428 Rank: 144/147 |
0.05586 (±0.00165) Rank: 133/147 |
0.10570 (±0.00121) Rank: 138/147 |
| mr | 6143.0 | 785.0 | 785.0 | 281.1 | 0.310 Rank: 147/147 |
0.424 Rank: 143/147 |
0.06733 (±0.00018) Rank: 137/147 |
0.11671 (±0.00132) Rank: 137/147 |
| psm | 6370.0 | 681.5 | 681.5 | 261.0 | 0.242 Rank: 142/147 |
0.151 Rank: 145/147 |
0.06130 (±0.00081) Rank: 135/147 |
0.13934 (±0.00112) Rank: 132/147 |
| sf | 6969.9 | 889.1 | 889.1 | 306.0 | 0.304 Rank: 147/147 |
0.367 Rank: 142/147 |
0.11974 (±0.00028) Rank: 136/147 |
0.20961 (±0.00072) Rank: 137/147 |
| spc | 5534.9 | 674.3 | 674.3 | 209.9 | 0.281 Rank: 142/147 |
0.340 Rank: 142/147 |
0.09122 (±0.00070) Rank: 134/147 |
0.16163 (±0.00031) Rank: 134/147 |
| avg | 5665.3 | 722.3 | 722.3 | 251.5 | 0.288 Rank: 142/147 |
0.360 Rank: 143/147 |
0.10794 (±0.00045) Rank: 136/147 |
0.18499 (±0.00022) Rank: 136/147 |
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: 0.41726 (±0.00175 over 3 run(s) / ±0.21452 over 9 scenes)
Rank (per category): 141 (of 147)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| bm | 4887.9 | 628.56 | 98.95 Rank: 136/147 |
3304.65 Rank: 46/147 |
3.351 Rank: 140/147 |
1.15292 Rank: 142/147 |
0.04373 (±0.00297) Rank: 141/147 |
0.09803 (±0.00578) Rank: 141/147 |
| fcs | 5450.1 | 812.74 | 96.97 Rank: 125/147 |
4968.38 Rank: 67/147 |
3.582 Rank: 135/147 |
0.39874 Rank: 138/147 |
0.45912 (±0.00336) Rank: 139/147 |
0.57744 (±0.00088) Rank: 137/147 |
| lms | 4539.5 | 733.23 | 98.19 Rank: 124/147 |
4304.94 Rank: 28/147 |
3.726 Rank: 136/147 |
0.36576 Rank: 135/147 |
0.63336 (±0.00235) Rank: 134/147 |
0.74984 (±0.00216) Rank: 134/147 |
| lb | 5132.1 | 721.69 | 97.90 Rank: 103/147 |
3749.93 Rank: 31/147 |
3.951 Rank: 135/147 |
0.63011 Rank: 139/147 |
0.30477 (±0.00623) Rank: 134/147 |
0.45372 (±0.00466) Rank: 135/147 |
| mc | 5960.5 | 717.71 | 97.24 Rank: 140/147 |
4160.40 Rank: 76/147 |
3.707 Rank: 140/147 |
0.68915 Rank: 145/147 |
0.23797 (±0.00429) Rank: 142/147 |
0.37221 (±0.00851) Rank: 142/147 |
| mr | 6143.0 | 790.22 | 91.68 Rank: 133/147 |
4224.57 Rank: 104/147 |
3.803 Rank: 130/147 |
0.79345 Rank: 143/147 |
0.16317 (±0.00138) Rank: 143/147 |
0.25078 (±0.00422) Rank: 142/147 |
| psm | 6370.0 | 677.46 | 88.08 Rank: 101/147 |
5652.98 Rank: 15/147 |
2.389 Rank: 141/147 |
0.93198 Rank: 145/147 |
0.04027 (±0.00367) Rank: 142/147 |
0.11114 (±0.00447) Rank: 142/147 |
| sf | 6969.9 | 883.37 | 94.06 Rank: 140/147 |
6690.79 Rank: 44/147 |
3.505 Rank: 141/147 |
0.39887 Rank: 135/147 |
0.50117 (±0.00629) Rank: 140/147 |
0.61526 (±0.00720) Rank: 141/147 |
| spc | 5534.9 | 695.09 | 98.01 Rank: 122/147 |
4199.94 Rank: 62/147 |
3.544 Rank: 137/147 |
0.90493 Rank: 141/147 |
0.38489 (±0.01185) Rank: 141/147 |
0.52691 (±0.01127) Rank: 139/147 |
| avg | 5665.3 | 740.01 | 95.68 Rank: 125/147 |
4584.07 Rank: 45/147 |
3.506 Rank: 138/147 |
0.69621 Rank: 142/147 |
0.30761 (±0.00116) Rank: 142/147 |
0.41726 (±0.00175) Rank: 141/147 |
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 —
