Submission ID: 00029
SuperPoint (2k features, NMS=4), DEGENSAC
Processed: 20-04-22. Download link: sid-00029-superpoint-2k-nms4.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: superpoint-nms4-r1200
- Descriptor: superpoint-nms4-r1200 (256 float32: 1024 bytes)
- Number of features: 2048
- Summary: SuperPoint (initial release based on Pytorch). Lowered detection threshold to obtain up to 2k features per image. Using the built-in Non-Maxima Suppression filter at 2 pixels. Images are resized to 1200 pixels on the largest side.
- Paper: https://arxiv.org/abs/1712.07629
- Website: https://github.com/MagicLeapResearch/SuperPointPretrainedNetwork
- Origin: Baseline
- Flags: is_baseline
Phototourism / Stereo track
mAA at 10 degrees: 0.28965 (±0.00063 over 3 run(s) / ±0.10509 over 9 scenes)
Rank (per category): 87 (of 108)
| Scene | Features | Matches (matcher) |
Matches (filter) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| bm | 2048.0 | 299.1 | 299.1 | 131.8 | 0.411 Rank: 90/108 |
0.815 Rank: 87/108 |
0.14107 (±0.00223) Rank: 56/108 |
0.22623 (±0.00152) Rank: 64/108 |
| fcs | 2048.0 | 293.9 | 293.9 | 134.0 | 0.373 Rank: 33/108 |
0.675 Rank: 93/108 |
0.31151 (±0.00281) Rank: 87/108 |
0.43006 (±0.00282) Rank: 87/108 |
| lms | 2048.0 | 296.5 | 296.5 | 136.9 | 0.369 Rank: 55/108 |
0.559 Rank: 91/108 |
0.33948 (±0.00266) Rank: 83/108 |
0.44551 (±0.00281) Rank: 83/108 |
| lb | 2048.0 | 303.9 | 303.9 | 122.7 | 0.347 Rank: 35/108 |
0.624 Rank: 83/108 |
0.25402 (±0.00223) Rank: 56/108 |
0.35102 (±0.00211) Rank: 72/108 |
| mc | 2048.0 | 283.0 | 283.0 | 112.8 | 0.380 Rank: 32/108 |
0.711 Rank: 91/108 |
0.12894 (±0.00223) Rank: 56/108 |
0.21569 (±0.00212) Rank: 87/108 |
| mr | 2048.0 | 281.8 | 281.8 | 123.8 | 0.408 Rank: 30/108 |
0.726 Rank: 94/108 |
0.11345 (±0.00119) Rank: 87/108 |
0.17206 (±0.00105) Rank: 87/108 |
| psm | 2048.0 | 186.1 | 186.1 | 64.2 | 0.292 Rank: 39/108 |
0.290 Rank: 91/108 |
0.06708 (±0.00056) Rank: 87/108 |
0.14131 (±0.00037) Rank: 86/108 |
| sf | 2048.0 | 360.8 | 360.8 | 177.2 | 0.353 Rank: 31/108 |
0.651 Rank: 90/108 |
0.25300 (±0.00205) Rank: 83/108 |
0.36378 (±0.00237) Rank: 85/108 |
| spc | 2048.0 | 277.7 | 277.7 | 100.8 | 0.339 Rank: 39/108 |
0.614 Rank: 94/108 |
0.16525 (±0.00131) Rank: 87/108 |
0.26124 (±0.00165) Rank: 89/108 |
| avg | 2048.0 | 287.0 | 287.0 | 122.7 | 0.364 Rank: 37/108 |
0.630 Rank: 91/108 |
0.19709 (±0.00030) Rank: 87/108 |
0.28965 (±0.00063) Rank: 87/108 |
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: 0.54663 (±0.00218 over 3 run(s) / ±0.16223 over 9 scenes)
Rank (per category): 81 (of 108)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| bm | 2048.0 | 178.14 | 98.80 Rank: 90/108 |
983.73 Rank: 91/108 |
4.557 Rank: 80/108 |
0.65033 Rank: 59/108 |
0.30787 (±0.00397) Rank: 59/108 |
0.44816 (±0.00245) Rank: 59/108 |
| fcs | 2048.0 | 185.30 | 93.36 Rank: 99/108 |
1457.10 Rank: 88/108 |
4.170 Rank: 66/108 |
0.37521 Rank: 93/108 |
0.55242 (±0.00206) Rank: 92/108 |
0.63641 (±0.00110) Rank: 91/108 |
| lms | 2048.0 | 177.22 | 95.47 Rank: 93/108 |
1194.66 Rank: 63/108 |
4.813 Rank: 44/108 |
0.32171 Rank: 58/108 |
0.71096 (±0.00760) Rank: 82/108 |
0.78433 (±0.00860) Rank: 82/108 |
| lb | 2048.0 | 211.25 | 97.57 Rank: 48/108 |
1085.76 Rank: 81/108 |
5.100 Rank: 14/108 |
0.51772 Rank: 55/108 |
0.53709 (±0.00081) Rank: 52/108 |
0.66178 (±0.00057) Rank: 51/108 |
| mc | 2048.0 | 152.17 | 95.15 Rank: 78/108 |
1096.61 Rank: 74/108 |
4.380 Rank: 44/108 |
0.60539 Rank: 95/108 |
0.33659 (±0.00610) Rank: 83/108 |
0.46984 (±0.00783) Rank: 81/108 |
| mr | 2048.0 | 164.24 | 87.79 Rank: 85/108 |
1034.82 Rank: 97/108 |
4.367 Rank: 38/108 |
0.67708 Rank: 90/108 |
0.25619 (±0.00269) Rank: 87/108 |
0.35720 (±0.00383) Rank: 85/108 |
| psm | 2048.0 | 78.76 | 75.14 Rank: 74/108 |
824.81 Rank: 96/108 |
2.946 Rank: 63/108 |
0.79907 Rank: 75/108 |
0.19064 (±0.00526) Rank: 64/108 |
0.26228 (±0.00672) Rank: 65/108 |
| sf | 2048.0 | 233.10 | 93.26 Rank: 88/108 |
1794.55 Rank: 58/108 |
4.491 Rank: 43/108 |
0.35065 Rank: 58/108 |
0.61861 (±0.00395) Rank: 75/108 |
0.70182 (±0.00348) Rank: 79/108 |
| spc | 2048.0 | 155.19 | 95.49 Rank: 93/108 |
1196.38 Rank: 89/108 |
4.153 Rank: 75/108 |
0.71244 Rank: 91/108 |
0.48117 (±0.00495) Rank: 91/108 |
0.59785 (±0.00419) Rank: 91/108 |
| avg | 2048.0 | 170.60 | 92.45 Rank: 81/108 |
1185.38 Rank: 90/108 |
4.331 Rank: 47/108 |
0.55662 Rank: 87/108 |
0.44350 (±0.00149) Rank: 81/108 |
0.54663 (±0.00218) Rank: 81/108 |
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 —
