Submission ID: 00641
MT-2-Hardnet-Pretraind-all-Datasets-512-bytes
Processed: 21-03-04. Download link: sid-00641-mt-2-hardnet-pretrained-all-datasets-magsac.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. MAGSAC 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: 0.34184 (±0.00032 over 3 run(s) / ±0.15471 over 9 scenes)
Rank (per category): 78 (of 108)
| Scene | Features | Matches (matcher) |
Matches (filter) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| bm | 2048.0 | 310.1 | 310.1 | 238.7 | 0.592 Rank: 3/108 |
0.734 Rank: 95/108 |
0.02255 (±0.00168) Rank: 95/108 |
0.05244 (±0.00172) Rank: 95/108 |
| fcs | 2048.0 | 261.2 | 261.2 | 195.6 | 0.374 Rank: 29/108 |
0.840 Rank: 76/108 |
0.44613 (±0.00109) Rank: 75/108 |
0.58606 (±0.00052) Rank: 71/108 |
| lms | 2048.0 | 340.2 | 340.2 | 258.0 | 0.381 Rank: 36/108 |
0.616 Rank: 83/108 |
0.35760 (±0.00161) Rank: 76/108 |
0.47117 (±0.00219) Rank: 76/108 |
| lb | 2048.0 | 265.9 | 265.9 | 174.5 | 0.409 Rank: 10/108 |
0.626 Rank: 82/108 |
0.18848 (±0.00168) Rank: 95/108 |
0.26561 (±0.00079) Rank: 89/108 |
| mc | 2048.0 | 286.0 | 286.0 | 211.0 | 0.509 Rank: 16/108 |
0.879 Rank: 60/108 |
0.24253 (±0.00168) Rank: 95/108 |
0.38055 (±0.00340) Rank: 60/108 |
| mr | 2048.0 | 373.6 | 373.6 | 308.1 | 0.479 Rank: 5/108 |
0.881 Rank: 76/108 |
0.19968 (±0.00138) Rank: 58/108 |
0.29188 (±0.00082) Rank: 58/108 |
| psm | 2048.0 | 153.2 | 153.2 | 104.4 | 0.379 Rank: 3/108 |
0.420 Rank: 54/108 |
0.08373 (±0.00076) Rank: 70/108 |
0.17388 (±0.00114) Rank: 55/108 |
| sf | 2048.0 | 376.1 | 376.1 | 303.2 | 0.449 Rank: 2/108 |
0.769 Rank: 78/108 |
0.32756 (±0.00081) Rank: 65/108 |
0.46366 (±0.00018) Rank: 65/108 |
| spc | 2048.0 | 263.3 | 263.3 | 184.0 | 0.479 Rank: 1/108 |
0.736 Rank: 80/108 |
0.24725 (±0.00203) Rank: 79/108 |
0.39132 (±0.00059) Rank: 77/108 |
| avg | 2048.0 | 292.2 | 292.2 | 219.7 | 0.450 Rank: 4/108 |
0.722 Rank: 81/108 |
0.23506 (±0.00055) Rank: 79/108 |
0.34184 (±0.00032) Rank: 78/108 |
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: 0.60963 (±0.00152 over 3 run(s) / ±0.18791 over 9 scenes)
Rank (per category): 53 (of 108)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| bm | 2048.0 | 308.98 | 99.22 Rank: 84/108 |
1478.01 Rank: 39/108 |
4.488 Rank: 87/108 |
0.87681 Rank: 92/108 |
0.15086 (±0.00419) Rank: 93/108 |
0.25544 (±0.00603) Rank: 93/108 |
| fcs | 2048.0 | 269.42 | 98.17 Rank: 13/108 |
1707.09 Rank: 62/108 |
4.032 Rank: 79/108 |
0.25719 Rank: 27/108 |
0.70427 (±0.01029) Rank: 34/108 |
0.78063 (±0.01320) Rank: 29/108 |
| lms | 2048.0 | 368.35 | 99.38 Rank: 9/108 |
1897.38 Rank: 27/108 |
4.467 Rank: 83/108 |
0.30359 Rank: 40/108 |
0.75461 (±0.01266) Rank: 50/108 |
0.83856 (±0.01240) Rank: 45/108 |
| lb | 2048.0 | 323.44 | 98.33 Rank: 30/108 |
1425.79 Rank: 40/108 |
4.520 Rank: 41/108 |
0.50434 Rank: 40/108 |
0.51116 (±0.00167) Rank: 59/108 |
0.62105 (±0.00326) Rank: 60/108 |
| mc | 2048.0 | 280.36 | 99.62 Rank: 31/108 |
1281.45 Rank: 46/108 |
4.571 Rank: 36/108 |
0.43719 Rank: 39/108 |
0.44981 (±0.00660) Rank: 40/108 |
0.60467 (±0.00827) Rank: 38/108 |
| mr | 2048.0 | 373.89 | 94.24 Rank: 15/108 |
1676.48 Rank: 40/108 |
4.466 Rank: 28/108 |
0.53290 Rank: 24/108 |
0.35694 (±0.01030) Rank: 37/108 |
0.48028 (±0.00875) Rank: 32/108 |
| psm | 2048.0 | 148.70 | 87.24 Rank: 36/108 |
1476.26 Rank: 46/108 |
3.039 Rank: 54/108 |
0.67800 Rank: 44/108 |
0.29778 (±0.00443) Rank: 39/108 |
0.38924 (±0.00414) Rank: 37/108 |
| sf | 2048.0 | 371.50 | 98.20 Rank: 25/108 |
1857.26 Rank: 48/108 |
4.559 Rank: 36/108 |
0.32333 Rank: 31/108 |
0.68306 (±0.00270) Rank: 38/108 |
0.78044 (±0.00315) Rank: 35/108 |
| spc | 2048.0 | 270.15 | 99.38 Rank: 36/108 |
1557.97 Rank: 49/108 |
4.359 Rank: 46/108 |
0.57822 Rank: 75/108 |
0.61772 (±0.00421) Rank: 63/108 |
0.73637 (±0.00560) Rank: 56/108 |
| avg | 2048.0 | 301.64 | 97.09 Rank: 32/108 |
1595.30 Rank: 43/108 |
4.278 Rank: 62/108 |
0.49907 Rank: 60/108 |
0.50291 (±0.00122) Rank: 55/108 |
0.60963 (±0.00152) Rank: 53/108 |
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 —
