Submission ID: 00640
MT-2-Hardnet-Pretraind-all-Datasets-512-bytes
Processed: 21-02-17. Download link: sid-00640-mt-2-hardnet-pretrained-all-datasets-degensac-check-on.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: Anonymous (to be released: 2020-6-12)
- Keypoint: mt-2
- Descriptor: hardnet (128 float32: 512 bytes)
- Number of features: 2048
- Summary: Local feature model learned via training with covariant constraint loss function. We take the top 2048 features by score. HardNet ,pre-trained on all datasets, is used as a descriptor head. DEGENSA with degeneracy check on is used as a robust estimator. Cyclic consistency matching with a threshold of 0.95 is used.
- Paper: Anonymous (to be released: 2020-6-12)
- Website: Anonymous (to be released: 2020-6-12)
- Origin: Submission
- Flags: is_submission
Phototourism / Stereo track
mAA at 10 degrees: 0.33122 (±0.00042 over 3 run(s) / ±0.15010 over 9 scenes)
Rank (per category): 82 (of 108)
| Scene | Features | Matches (matcher) |
Matches (filter) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| bm | 2048.0 | 307.3 | 307.3 | 156.2 | 0.592 Rank: 3/108 |
0.759 Rank: 94/108 |
0.02417 (±0.00066) Rank: 94/108 |
0.05489 (±0.00053) Rank: 94/108 |
| fcs | 2048.0 | 252.9 | 252.9 | 119.8 | 0.374 Rank: 29/108 |
0.860 Rank: 61/108 |
0.43129 (±0.00298) Rank: 80/108 |
0.57278 (±0.00196) Rank: 76/108 |
| lms | 2048.0 | 331.1 | 331.1 | 151.2 | 0.381 Rank: 36/108 |
0.650 Rank: 78/108 |
0.35417 (±0.00197) Rank: 80/108 |
0.47393 (±0.00208) Rank: 75/108 |
| lb | 2048.0 | 258.8 | 258.8 | 100.7 | 0.409 Rank: 10/108 |
0.643 Rank: 78/108 |
0.18316 (±0.00066) Rank: 94/108 |
0.25555 (±0.00170) Rank: 91/108 |
| mc | 2048.0 | 272.4 | 272.4 | 124.4 | 0.509 Rank: 16/108 |
0.902 Rank: 45/108 |
0.23270 (±0.00066) Rank: 94/108 |
0.36877 (±0.00486) Rank: 65/108 |
| mr | 2048.0 | 341.2 | 341.2 | 168.0 | 0.479 Rank: 5/108 |
0.913 Rank: 49/108 |
0.19366 (±0.00102) Rank: 59/108 |
0.28282 (±0.00084) Rank: 60/108 |
| psm | 2048.0 | 150.8 | 150.8 | 61.1 | 0.379 Rank: 3/108 |
0.423 Rank: 50/108 |
0.08093 (±0.00036) Rank: 75/108 |
0.16811 (±0.00173) Rank: 62/108 |
| sf | 2048.0 | 349.9 | 349.9 | 169.6 | 0.449 Rank: 2/108 |
0.787 Rank: 62/108 |
0.29372 (±0.00281) Rank: 77/108 |
0.43007 (±0.00282) Rank: 78/108 |
| spc | 2048.0 | 256.3 | 256.3 | 107.9 | 0.479 Rank: 1/108 |
0.762 Rank: 78/108 |
0.23498 (±0.00325) Rank: 81/108 |
0.37407 (±0.00352) Rank: 81/108 |
| avg | 2048.0 | 280.1 | 280.1 | 128.8 | 0.450 Rank: 4/108 |
0.744 Rank: 77/108 |
0.22542 (±0.00063) Rank: 81/108 |
0.33122 (±0.00042) Rank: 82/108 |
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: 0.60128 (±0.00311 over 3 run(s) / ±0.18751 over 9 scenes)
Rank (per category): 60 (of 108)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| bm | 2048.0 | 306.17 | 99.39 Rank: 80/108 |
1464.66 Rank: 41/108 |
4.496 Rank: 85/108 |
0.86791 Rank: 91/108 |
0.15153 (±0.00197) Rank: 91/108 |
0.25840 (±0.00459) Rank: 91/108 |
| fcs | 2048.0 | 260.85 | 97.70 Rank: 35/108 |
1656.49 Rank: 68/108 |
3.997 Rank: 83/108 |
0.28519 Rank: 62/108 |
0.68375 (±0.01032) Rank: 49/108 |
0.75739 (±0.01175) Rank: 49/108 |
| lms | 2048.0 | 358.98 | 99.22 Rank: 20/108 |
1860.10 Rank: 30/108 |
4.466 Rank: 84/108 |
0.29064 Rank: 15/108 |
0.75836 (±0.00997) Rank: 47/108 |
0.83920 (±0.01099) Rank: 44/108 |
| lb | 2048.0 | 314.70 | 98.03 Rank: 40/108 |
1388.26 Rank: 44/108 |
4.520 Rank: 42/108 |
0.53337 Rank: 73/108 |
0.48639 (±0.00977) Rank: 64/108 |
0.59490 (±0.00943) Rank: 71/108 |
| mc | 2048.0 | 266.90 | 99.59 Rank: 32/108 |
1238.19 Rank: 52/108 |
4.583 Rank: 34/108 |
0.42140 Rank: 35/108 |
0.45247 (±0.00386) Rank: 39/108 |
0.60608 (±0.00401) Rank: 37/108 |
| mr | 2048.0 | 342.12 | 94.46 Rank: 9/108 |
1585.16 Rank: 51/108 |
4.539 Rank: 20/108 |
0.56814 Rank: 64/108 |
0.36175 (±0.00686) Rank: 31/108 |
0.48868 (±0.00765) Rank: 27/108 |
| psm | 2048.0 | 146.29 | 86.01 Rank: 40/108 |
1401.91 Rank: 49/108 |
3.012 Rank: 57/108 |
0.68075 Rank: 45/108 |
0.27264 (±0.00643) Rank: 47/108 |
0.36116 (±0.00931) Rank: 41/108 |
| sf | 2048.0 | 346.30 | 98.16 Rank: 26/108 |
1772.60 Rank: 64/108 |
4.580 Rank: 33/108 |
0.35380 Rank: 61/108 |
0.68148 (±0.00302) Rank: 40/108 |
0.78083 (±0.00310) Rank: 34/108 |
| spc | 2048.0 | 262.88 | 99.22 Rank: 43/108 |
1525.99 Rank: 56/108 |
4.362 Rank: 45/108 |
0.58093 Rank: 76/108 |
0.60328 (±0.00350) Rank: 67/108 |
0.72487 (±0.00297) Rank: 63/108 |
| avg | 2048.0 | 289.47 | 96.86 Rank: 36/108 |
1543.70 Rank: 47/108 |
4.284 Rank: 60/108 |
0.50913 Rank: 64/108 |
0.49463 (±0.00304) Rank: 58/108 |
0.60128 (±0.00311) Rank: 60/108 |
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 —
