Submission ID: 00045
D2-Net (multi-scale), DEGENSAC
Processed: 20-04-22. Download link: sid-00045-d2net-multiscale-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-multiscale
- Descriptor: d2net-multiscale (512 float32: 2048 bytes)
- Number of features: 8000
- Summary: D2-Net, multi-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: 0.20486 (±0.00062 over 3 run(s) / ±0.11091 over 9 scenes)
Rank (per category): 135 (of 147)
| Scene | Features | Matches (matcher) |
Matches (filter) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| bm | 5998.0 | 603.8 | 603.8 | 293.9 | 0.323 Rank: 141/147 |
0.473 Rank: 140/147 |
0.02673 (±0.00067) Rank: 134/147 |
0.05611 (±0.00076) Rank: 135/147 |
| fcs | 7158.4 | 830.0 | 830.0 | 385.9 | 0.301 Rank: 141/147 |
0.420 Rank: 141/147 |
0.20924 (±0.00283) Rank: 136/147 |
0.35319 (±0.00216) Rank: 135/147 |
| lms | 5762.2 | 769.1 | 769.1 | 383.3 | 0.342 Rank: 141/147 |
0.402 Rank: 139/147 |
0.27632 (±0.00321) Rank: 133/147 |
0.42487 (±0.00245) Rank: 129/147 |
| lb | 6639.2 | 632.2 | 632.2 | 270.3 | 0.320 Rank: 136/147 |
0.355 Rank: 140/147 |
0.12282 (±0.00067) Rank: 134/147 |
0.20891 (±0.00288) Rank: 133/147 |
| mc | 7229.6 | 698.1 | 698.1 | 330.4 | 0.354 Rank: 141/147 |
0.453 Rank: 142/147 |
0.05584 (±0.00067) Rank: 134/147 |
0.10613 (±0.00187) Rank: 137/147 |
| mr | 7536.2 | 840.4 | 840.4 | 429.1 | 0.357 Rank: 141/147 |
0.497 Rank: 141/147 |
0.07209 (±0.00187) Rank: 135/147 |
0.12890 (±0.00250) Rank: 135/147 |
| psm | 7312.9 | 625.1 | 625.1 | 310.8 | 0.269 Rank: 141/147 |
0.161 Rank: 144/147 |
0.07214 (±0.00024) Rank: 130/147 |
0.15652 (±0.00076) Rank: 128/147 |
| sf | 7689.2 | 822.2 | 822.2 | 374.4 | 0.315 Rank: 146/147 |
0.364 Rank: 143/147 |
0.11967 (±0.00153) Rank: 137/147 |
0.21695 (±0.00093) Rank: 136/147 |
| spc | 6991.6 | 629.2 | 629.2 | 271.2 | 0.312 Rank: 141/147 |
0.345 Rank: 141/147 |
0.10465 (±0.00044) Rank: 133/147 |
0.19216 (±0.00092) Rank: 133/147 |
| avg | 6924.1 | 716.7 | 716.7 | 338.8 | 0.321 Rank: 141/147 |
0.386 Rank: 141/147 |
0.11772 (±0.00051) Rank: 135/147 |
0.20486 (±0.00062) Rank: 135/147 |
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: 0.42782 (±0.00544 over 3 run(s) / ±0.21858 over 9 scenes)
Rank (per category): 139 (of 147)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| bm | 5998.0 | 607.80 | 98.99 Rank: 135/147 |
3596.18 Rank: 29/147 |
3.349 Rank: 141/147 |
1.12978 Rank: 141/147 |
0.03778 (±0.00151) Rank: 142/147 |
0.08838 (±0.00246) Rank: 142/147 |
| fcs | 7158.4 | 862.76 | 97.81 Rank: 87/147 |
6845.91 Rank: 14/147 |
3.342 Rank: 139/147 |
0.38312 Rank: 135/147 |
0.46630 (±0.00466) Rank: 138/147 |
0.59499 (±0.00770) Rank: 136/147 |
| lms | 5762.2 | 825.98 | 99.33 Rank: 40/147 |
5934.66 Rank: 4/147 |
3.596 Rank: 138/147 |
0.32411 Rank: 114/147 |
0.64998 (±0.00357) Rank: 133/147 |
0.77364 (±0.00660) Rank: 133/147 |
| lb | 6639.2 | 747.73 | 98.14 Rank: 95/147 |
4749.19 Rank: 12/147 |
3.625 Rank: 140/147 |
0.59672 Rank: 135/147 |
0.30093 (±0.01479) Rank: 135/147 |
0.45910 (±0.01497) Rank: 134/147 |
| mc | 7229.6 | 688.94 | 99.08 Rank: 124/147 |
5095.78 Rank: 22/147 |
3.466 Rank: 143/147 |
0.64525 Rank: 142/147 |
0.22461 (±0.00825) Rank: 143/147 |
0.36744 (±0.01162) Rank: 143/147 |
| mr | 7536.2 | 839.60 | 94.89 Rank: 37/147 |
6434.74 Rank: 17/147 |
3.515 Rank: 141/147 |
0.75534 Rank: 140/147 |
0.17104 (±0.01013) Rank: 141/147 |
0.26999 (±0.01224) Rank: 140/147 |
| psm | 7312.9 | 625.39 | 89.24 Rank: 91/147 |
5963.17 Rank: 13/147 |
2.442 Rank: 140/147 |
0.88890 Rank: 70/147 |
0.05152 (±0.00175) Rank: 140/147 |
0.13249 (±0.00267) Rank: 140/147 |
| sf | 7689.2 | 815.62 | 96.65 Rank: 123/147 |
7469.07 Rank: 17/147 |
3.288 Rank: 143/147 |
0.41199 Rank: 140/147 |
0.48354 (±0.00434) Rank: 142/147 |
0.61973 (±0.00448) Rank: 140/147 |
| spc | 6991.6 | 653.13 | 98.59 Rank: 103/147 |
5183.86 Rank: 22/147 |
3.308 Rank: 143/147 |
0.80714 Rank: 138/147 |
0.39023 (±0.00172) Rank: 139/147 |
0.54460 (±0.00309) Rank: 138/147 |
| avg | 6924.1 | 740.77 | 96.97 Rank: 89/147 |
5696.95 Rank: 15/147 |
3.326 Rank: 143/147 |
0.66026 Rank: 139/147 |
0.30844 (±0.00380) Rank: 141/147 |
0.42782 (±0.00544) Rank: 139/147 |
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 —
