Submission ID: 00046
D2-Net (single-scale), DEGENSAC
Processed: 20-04-23. Download link: sid-00046-d2net-singlescale-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-singlescale
- Descriptor: d2net-singlescale (512 float32: 2048 bytes)
- Number of features: 2048
- Summary: D2-Net, single-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: 0.16324 (±0.00082 over 3 run(s) / ±0.08942 over 9 scenes)
Rank (per category): 97 (of 108)
| Scene | Features | Matches (matcher) |
Matches (filter) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| bm | 2047.8 | 327.9 | 327.9 | 143.0 | 0.258 Rank: 99/108 |
0.423 Rank: 100/108 |
0.02247 (±0.00077) Rank: 96/108 |
0.04886 (±0.00071) Rank: 97/108 |
| fcs | 2045.8 | 409.1 | 409.1 | 174.2 | 0.188 Rank: 99/108 |
0.395 Rank: 100/108 |
0.16878 (±0.00176) Rank: 96/108 |
0.29134 (±0.00167) Rank: 96/108 |
| lms | 2015.3 | 420.8 | 420.8 | 192.2 | 0.235 Rank: 99/108 |
0.291 Rank: 101/108 |
0.20954 (±0.00224) Rank: 95/108 |
0.32239 (±0.00091) Rank: 94/108 |
| lb | 2045.5 | 359.0 | 359.0 | 144.1 | 0.209 Rank: 99/108 |
0.313 Rank: 100/108 |
0.12239 (±0.00077) Rank: 96/108 |
0.20417 (±0.00172) Rank: 93/108 |
| mc | 2044.7 | 301.8 | 301.8 | 115.4 | 0.210 Rank: 99/108 |
0.418 Rank: 100/108 |
0.03139 (±0.00077) Rank: 96/108 |
0.06651 (±0.00085) Rank: 98/108 |
| mr | 2033.0 | 356.7 | 356.7 | 142.9 | 0.202 Rank: 99/108 |
0.408 Rank: 100/108 |
0.05149 (±0.00157) Rank: 97/108 |
0.09666 (±0.00327) Rank: 98/108 |
| psm | 2030.8 | 299.0 | 299.0 | 144.0 | 0.150 Rank: 99/108 |
0.112 Rank: 100/108 |
0.05155 (±0.00084) Rank: 99/108 |
0.12885 (±0.00242) Rank: 92/108 |
| sf | 2044.1 | 373.6 | 373.6 | 134.8 | 0.158 Rank: 100/108 |
0.332 Rank: 100/108 |
0.07297 (±0.00091) Rank: 97/108 |
0.14177 (±0.00207) Rank: 97/108 |
| spc | 2037.0 | 355.7 | 355.7 | 137.3 | 0.209 Rank: 99/108 |
0.340 Rank: 100/108 |
0.09317 (±0.00236) Rank: 95/108 |
0.16865 (±0.00150) Rank: 96/108 |
| avg | 2038.2 | 356.0 | 356.0 | 147.5 | 0.202 Rank: 99/108 |
0.337 Rank: 100/108 |
0.09153 (±0.00052) Rank: 97/108 |
0.16324 (±0.00082) Rank: 97/108 |
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: 0.36911 (±0.00167 over 3 run(s) / ±0.22314 over 9 scenes)
Rank (per category): 98 (of 108)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| bm | 2047.8 | 328.12 | 97.68 Rank: 94/108 |
1687.38 Rank: 28/108 |
3.266 Rank: 100/108 |
1.22965 Rank: 100/108 |
0.03328 (±0.00224) Rank: 98/108 |
0.07646 (±0.00439) Rank: 98/108 |
| fcs | 2045.8 | 418.69 | 96.71 Rank: 56/108 |
2386.33 Rank: 18/108 |
3.568 Rank: 96/108 |
0.38998 Rank: 94/108 |
0.43616 (±0.00786) Rank: 95/108 |
0.56344 (±0.01165) Rank: 94/108 |
| lms | 2015.3 | 444.30 | 97.36 Rank: 81/108 |
2737.28 Rank: 1/108 |
3.777 Rank: 98/108 |
0.32453 Rank: 62/108 |
0.58905 (±0.00390) Rank: 95/108 |
0.71628 (±0.00357) Rank: 92/108 |
| lb | 2045.5 | 398.87 | 97.93 Rank: 42/108 |
2181.67 Rank: 8/108 |
3.762 Rank: 96/108 |
0.56581 Rank: 86/108 |
0.29766 (±0.00942) Rank: 94/108 |
0.45540 (±0.00994) Rank: 93/108 |
| mc | 2044.7 | 302.73 | 93.38 Rank: 92/108 |
1401.29 Rank: 36/108 |
3.369 Rank: 99/108 |
0.80554 Rank: 100/108 |
0.14504 (±0.00421) Rank: 98/108 |
0.25082 (±0.00618) Rank: 98/108 |
| mr | 2033.0 | 360.55 | 87.49 Rank: 88/108 |
1378.89 Rank: 74/108 |
3.752 Rank: 90/108 |
0.84305 Rank: 97/108 |
0.11045 (±0.00176) Rank: 98/108 |
0.18294 (±0.00109) Rank: 98/108 |
| psm | 2030.8 | 293.50 | 75.71 Rank: 73/108 |
1982.13 Rank: 27/108 |
2.237 Rank: 99/108 |
0.96732 Rank: 100/108 |
0.01248 (±0.00159) Rank: 101/108 |
0.04503 (±0.00218) Rank: 99/108 |
| sf | 2044.1 | 371.43 | 92.39 Rank: 95/108 |
2148.99 Rank: 22/108 |
3.369 Rank: 99/108 |
0.46546 Rank: 97/108 |
0.39653 (±0.00761) Rank: 98/108 |
0.51177 (±0.01102) Rank: 98/108 |
| spc | 2037.0 | 363.95 | 97.54 Rank: 73/108 |
1960.83 Rank: 24/108 |
3.612 Rank: 98/108 |
0.89309 Rank: 98/108 |
0.37727 (±0.00146) Rank: 97/108 |
0.51987 (±0.00458) Rank: 96/108 |
| avg | 2038.2 | 364.68 | 92.91 Rank: 80/108 |
1984.98 Rank: 21/108 |
3.412 Rank: 98/108 |
0.72049 Rank: 98/108 |
0.26644 (±0.00113) Rank: 98/108 |
0.36911 (±0.00167) Rank: 98/108 |
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 —
