Submission ID: 00512
Hardnet with DEGENSAC for stereo
Processed: 20-04-25. Download link: sid-00512-siftdef-hardnet-64.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: Vu Trung Nghia & Nguyen Trung Hieu (contact)
- Keypoint: siftdef
- Descriptor: hardnet-64 (128 float32: 512 bytes)
- Number of features: 2048
- Summary: Using hardnet network to embed a patch (32 x 32) to a 128 (float32) dimensions vector, For stereo, we use DEGENSAC (Chum et al, CVPR'05) with optimal settings
- Paper: N/A
- Website: N/A
- Origin: Submission
- Flags: is_submission, is_challenge_2020
Phototourism / Stereo track
mAA at 10 degrees: 0.20771 (±0.00049 over 3 run(s) / ±0.09905 over 9 scenes)
Rank (per category): 93 (of 108)
| Scene | Features | Matches (matcher) |
Matches (filter) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| bm | 2000.0 | 98.6 | 98.6 | 55.7 | 0.193 Rank: 100/108 |
0.655 Rank: 98/108 |
0.02805 (±0.00047) Rank: 92/108 |
0.06355 (±0.00070) Rank: 92/108 |
| fcs | 2000.0 | 116.8 | 116.8 | 58.3 | 0.163 Rank: 100/108 |
0.649 Rank: 97/108 |
0.25476 (±0.00378) Rank: 92/108 |
0.36384 (±0.00259) Rank: 92/108 |
| lms | 2000.0 | 99.4 | 99.4 | 46.3 | 0.177 Rank: 101/108 |
0.469 Rank: 98/108 |
0.22192 (±0.00248) Rank: 94/108 |
0.31437 (±0.00260) Rank: 95/108 |
| lb | 2000.0 | 95.5 | 95.5 | 45.0 | 0.177 Rank: 100/108 |
0.496 Rank: 97/108 |
0.12303 (±0.00047) Rank: 92/108 |
0.17850 (±0.00125) Rank: 97/108 |
| mc | 2000.0 | 112.0 | 112.0 | 58.8 | 0.191 Rank: 100/108 |
0.697 Rank: 94/108 |
0.11704 (±0.00047) Rank: 92/108 |
0.20642 (±0.00156) Rank: 91/108 |
| mr | 1995.5 | 103.6 | 103.6 | 57.5 | 0.164 Rank: 101/108 |
0.635 Rank: 99/108 |
0.09717 (±0.00117) Rank: 90/108 |
0.15421 (±0.00188) Rank: 92/108 |
| psm | 2000.0 | 66.4 | 66.4 | 35.8 | 0.141 Rank: 100/108 |
0.272 Rank: 95/108 |
0.04347 (±0.00063) Rank: 103/108 |
0.10097 (±0.00085) Rank: 103/108 |
| sf | 2000.0 | 126.9 | 126.9 | 67.4 | 0.167 Rank: 99/108 |
0.644 Rank: 92/108 |
0.21929 (±0.00025) Rank: 89/108 |
0.32759 (±0.00159) Rank: 89/108 |
| spc | 2000.0 | 93.9 | 93.9 | 42.4 | 0.173 Rank: 101/108 |
0.506 Rank: 99/108 |
0.09237 (±0.00188) Rank: 96/108 |
0.15992 (±0.00198) Rank: 97/108 |
| avg | 1999.5 | 101.5 | 101.5 | 51.9 | 0.172 Rank: 100/108 |
0.558 Rank: 98/108 |
0.13301 (±0.00049) Rank: 93/108 |
0.20771 (±0.00049) Rank: 93/108 |
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: 0.38463 (±0.00275 over 3 run(s) / ±0.18874 over 9 scenes)
Rank (per category): 96 (of 108)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| bm | 2000.0 | 172.63 | 95.95 Rank: 99/108 |
1398.38 Rank: 42/108 |
3.145 Rank: 101/108 |
1.15310 Rank: 96/108 |
0.03332 (±0.00134) Rank: 97/108 |
0.07696 (±0.00180) Rank: 97/108 |
| fcs | 2000.0 | 200.33 | 91.72 Rank: 101/108 |
1910.60 Rank: 40/108 |
3.056 Rank: 100/108 |
0.40960 Rank: 95/108 |
0.42916 (±0.01549) Rank: 96/108 |
0.52697 (±0.01869) Rank: 96/108 |
| lms | 2000.0 | 188.21 | 94.76 Rank: 95/108 |
1856.62 Rank: 31/108 |
3.163 Rank: 101/108 |
0.36923 Rank: 85/108 |
0.58074 (±0.02561) Rank: 96/108 |
0.68103 (±0.02711) Rank: 96/108 |
| lb | 2000.0 | 194.18 | 95.19 Rank: 85/108 |
1612.50 Rank: 29/108 |
3.245 Rank: 100/108 |
0.61236 Rank: 99/108 |
0.29520 (±0.00603) Rank: 95/108 |
0.41566 (±0.00716) Rank: 95/108 |
| mc | 2000.0 | 188.81 | 93.26 Rank: 95/108 |
1794.62 Rank: 23/108 |
3.183 Rank: 100/108 |
0.64779 Rank: 97/108 |
0.25823 (±0.00816) Rank: 95/108 |
0.38537 (±0.00958) Rank: 94/108 |
| mr | 1995.5 | 167.93 | 83.07 Rank: 101/108 |
1308.32 Rank: 82/108 |
3.172 Rank: 100/108 |
0.83096 Rank: 96/108 |
0.15524 (±0.01266) Rank: 97/108 |
0.23467 (±0.01821) Rank: 96/108 |
| psm | 2000.0 | 139.76 | 73.05 Rank: 84/108 |
1409.25 Rank: 48/108 |
2.375 Rank: 97/108 |
0.85523 Rank: 85/108 |
0.08512 (±0.00410) Rank: 89/108 |
0.15294 (±0.00803) Rank: 88/108 |
| sf | 2000.0 | 201.90 | 91.57 Rank: 98/108 |
2096.45 Rank: 27/108 |
3.153 Rank: 100/108 |
0.40122 Rank: 88/108 |
0.48627 (±0.00643) Rank: 96/108 |
0.58559 (±0.00726) Rank: 96/108 |
| spc | 2000.0 | 179.55 | 93.43 Rank: 100/108 |
1846.92 Rank: 28/108 |
2.961 Rank: 101/108 |
0.97525 Rank: 100/108 |
0.27867 (±0.00537) Rank: 99/108 |
0.40245 (±0.00441) Rank: 100/108 |
| avg | 1999.5 | 181.48 | 90.22 Rank: 97/108 |
1692.63 Rank: 37/108 |
3.050 Rank: 100/108 |
0.69497 Rank: 95/108 |
0.28911 (±0.00275) Rank: 97/108 |
0.38463 (±0.00275) Rank: 96/108 |
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 —
