Submission ID: 7bd5e885
mss
Processed: 21-05-24. Download link: 7bd5e885d0a8c4ef-520json.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.
- Phototourism dataset: Stereo track / Multiview track
- Prague Parks dataset: Stereo track / Multiview track
- Google Urban dataset: Stereo track / Multiview track
Metadata
- Authors: Megvii 3D (contact)
- Keypoint: pt
- Descriptor: pt (128 float32: 512 bytes)
- Number of features: 2048
- Summary: pt
- Paper: N/A
- Website: N/A
- Processing date: 21-05-24
Phototourism dataset / Stereo track
mAA at 10 degrees: 0.58689 (±0.00000 over 1 run(s) / ±0.13234 over 9 scenes)
Rank (per category): 17 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
BM | 2048.0 | — | 374.8 | 0.460 Rank: 46/87 |
0.897 Rank: 34/87 |
0.35114 (±0.00000) Rank: 12/87 |
0.50513 (±0.00000) Rank: 12/87 |
FCS | 2048.0 | — | 530.6 | 0.417 Rank: 18/87 |
0.872 Rank: 30/87 |
0.65573 (±0.00000) Rank: 20/87 |
0.77198 (±0.00000) Rank: 23/87 |
LMS | 2048.0 | — | 547.5 | 0.448 Rank: 27/87 |
0.690 Rank: 31/87 |
0.62271 (±0.00000) Rank: 18/87 |
0.74372 (±0.00000) Rank: 18/87 |
LB | 2047.7 | — | 398.0 | 0.402 Rank: 30/87 |
0.702 Rank: 46/87 |
0.49219 (±0.00000) Rank: 12/87 |
0.60592 (±0.00000) Rank: 25/87 |
MC | 2048.0 | — | 416.1 | 0.419 Rank: 37/87 |
0.857 Rank: 49/87 |
0.34723 (±0.00000) Rank: 12/87 |
0.50414 (±0.00000) Rank: 29/87 |
MR | 2048.0 | — | 446.6 | 0.425 Rank: 41/87 |
0.901 Rank: 29/87 |
0.28703 (±0.00000) Rank: 9/87 |
0.40228 (±0.00000) Rank: 10/87 |
PSM | 2048.0 | — | 362.4 | 0.329 Rank: 55/87 |
0.611 Rank: 32/87 |
0.24365 (±0.00000) Rank: 18/87 |
0.39975 (±0.00000) Rank: 23/87 |
SF | 2048.0 | — | 517.1 | 0.376 Rank: 33/87 |
0.789 Rank: 39/87 |
0.51855 (±0.00000) Rank: 17/87 |
0.66035 (±0.00000) Rank: 17/87 |
SPC | 2048.0 | — | 418.2 | 0.383 Rank: 46/87 |
0.800 Rank: 34/87 |
0.52343 (±0.00000) Rank: 18/87 |
0.68875 (±0.00000) Rank: 16/87 |
Avg | 2048.0 | — | 445.7 | 0.407 Rank: 43/87 |
0.791 Rank: 32/87 |
0.44907 (±0.00000) Rank: 17/87 |
0.58689 (±0.00000) Rank: 17/87 |
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 dataset / Multiview track
mAA at 10 degrees: N/A (±0.00000 over 1 run(s) / ±0.10078 over 9 scenes)
Rank (per category): 17 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
BM | 2048.0 | 373.81 | 99.89 Rank: 29/87 |
1981.32 Rank: 36/87 |
4.902 Rank: 45/87 |
0.34272 Rank: 14/87 |
0.59182 (±0.00000) Rank: 20/87 |
0.72625 (±0.00000) Rank: 21/87 |
FCS | 2048.0 | 538.79 | 98.20 Rank: 25/87 |
2659.75 Rank: 27/87 |
5.119 Rank: 19/87 |
0.22598 Rank: 8/87 |
0.76028 (±0.00000) Rank: 9/87 |
0.81851 (±0.00000) Rank: 7/87 |
LMS | 2048.0 | 574.22 | 99.28 Rank: 35/87 |
2005.96 Rank: 28/87 |
5.837 Rank: 23/87 |
0.29685 Rank: 43/87 |
0.87058 (±0.00000) Rank: 6/87 |
0.91838 (±0.00000) Rank: 6/87 |
LB | 2047.7 | 475.51 | 98.23 Rank: 32/87 |
2104.48 Rank: 29/87 |
5.358 Rank: 27/87 |
0.49744 Rank: 54/87 |
0.72150 (±0.00000) Rank: 22/87 |
0.81262 (±0.00000) Rank: 23/87 |
MC | 2048.0 | 409.56 | 100.00 Rank: 1/87 |
2106.72 Rank: 41/87 |
5.146 Rank: 29/87 |
0.35530 Rank: 29/87 |
0.54928 (±0.00000) Rank: 35/87 |
0.69437 (±0.00000) Rank: 35/87 |
MR | 2048.0 | 440.41 | 95.33 Rank: 16/87 |
2237.48 Rank: 35/87 |
4.923 Rank: 16/87 |
0.49884 Rank: 21/87 |
0.43977 (±0.00000) Rank: 19/87 |
0.55856 (±0.00000) Rank: 24/87 |
PSM | 2048.0 | 360.33 | 99.27 Rank: 28/87 |
2418.83 Rank: 35/87 |
4.589 Rank: 17/87 |
0.32548 Rank: 14/87 |
0.65457 (±0.00000) Rank: 12/87 |
0.74288 (±0.00000) Rank: 13/87 |
SF | 2048.0 | 509.37 | 100.00 Rank: 1/87 |
2810.59 Rank: 27/87 |
5.107 Rank: 6/87 |
0.27358 Rank: 3/87 |
0.78699 (±0.00000) Rank: 3/87 |
0.86921 (±0.00000) Rank: 2/87 |
SPC | 2048.0 | 428.68 | 100.00 Rank: 1/87 |
2280.95 Rank: 36/87 |
5.172 Rank: 29/87 |
0.46793 Rank: 45/87 |
0.72826 (±0.00000) Rank: 37/87 |
0.81908 (±0.00000) Rank: 45/87 |
Avg | 2048.0 | 456.74 | 98.91 Rank: 12/87 |
2289.56 Rank: 34/87 |
5.128 Rank: 23/87 |
0.36490 Rank: 22/87 |
0.67812 (±0.00000) Rank: 16/87 |
0.77332 (±0.00000) Rank: 17/87 |
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 10 images. 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 —
Prague Parks dataset / Stereo track
mAA at 10 degrees: 0.74055 (±0.00000 over 1 run(s) / ±0.04779 over 3 scenes)
Rank (per category): 29 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
Lizard | 2048.0 | — | 301.2 | 0.035 Rank: 32/87 |
0.012 Rank: 14/87 |
0.58141 (±0.00000) Rank: 24/87 |
0.71983 (±0.00000) Rank: 22/87 |
Pond | 2048.0 | — | 295.5 | 0.048 Rank: 41/87 |
0.049 Rank: 28/87 |
0.55591 (±0.00000) Rank: 29/87 |
0.69519 (±0.00000) Rank: 31/87 |
Tree | 2048.0 | — | 190.0 | 0.029 Rank: 39/87 |
0.025 Rank: 12/87 |
0.69051 (±0.00000) Rank: 25/87 |
0.80662 (±0.00000) Rank: 24/87 |
Avg | 2048.0 | — | 262.2 | 0.038 Rank: 36/87 |
0.029 Rank: 8/87 |
0.60928 (±0.00000) Rank: 23/87 |
0.74055 (±0.00000) Rank: 29/87 |
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.
— Lizard —
— Pond —
— Tree —
Prague Parks dataset / Multiview track
mAA at 10 degrees: N/A (±0.00000 over 1 run(s) / ±0.14702 over 3 scenes)
Rank (per category): 14 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
Lizard | 2048.0 | 319.92 | 85.73 Rank: 1/87 |
1027.54 Rank: 34/87 |
3.472 Rank: 7/87 |
14.25999 Rank: 50/87 |
0.58889 (±0.00000) Rank: 20/87 |
0.66966 (±0.00000) Rank: 29/87 |
Pond | 2048.0 | 339.04 | 69.60 Rank: 18/87 |
862.89 Rank: 40/87 |
2.970 Rank: 24/87 |
0.42249 Rank: 24/87 |
0.27902 (±0.00000) Rank: 37/87 |
0.30972 (±0.00000) Rank: 40/87 |
Tree | 2048.0 | 185.08 | 72.73 Rank: 5/87 |
487.03 Rank: 39/87 |
3.073 Rank: 9/87 |
6.07469 Rank: 42/87 |
0.46840 (±0.00000) Rank: 9/87 |
0.50013 (±0.00000) Rank: 10/87 |
Avg | 2048.0 | 281.34 | 76.02 Rank: 3/87 |
792.49 Rank: 38/87 |
3.172 Rank: 11/87 |
6.91905 Rank: 43/87 |
0.44544 (±0.00000) Rank: 11/87 |
0.49317 (±0.00000) Rank: 14/87 |
Google Urban dataset / Stereo track
mAA at 10 degrees: 0.40347 (±0.00000 over 1 run(s) / ±0.13889 over 17 scenes)
Rank (per category): 17 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
mAA(5o) | mAA(10o) |
Amsterdam (AMS) | 2048.0 | — | 286.5 | 0.26786 (±0.00000) Rank: 9/87 |
0.42723 (±0.00000) Rank: 8/87 |
Bangkok (BGK) | 2048.0 | — | 411.3 | 0.06203 (±0.00000) Rank: 22/87 |
0.18164 (±0.00000) Rank: 16/87 |
Barcelona (BCN) | 2046.3 | — | 292.0 | 0.11355 (±0.00000) Rank: 28/87 |
0.21062 (±0.00000) Rank: 19/87 |
Buenos Aires (BAR) | 2048.0 | — | 281.6 | 0.22831 (±0.00000) Rank: 9/87 |
0.36804 (±0.00000) Rank: 27/87 |
Cambridge (CAM) | 2047.6 | — | 290.9 | 0.29863 (±0.00000) Rank: 9/87 |
0.45171 (±0.00000) Rank: 26/87 |
Cannes (CAN) | 2048.0 | — | 351.0 | 0.39539 (±0.00000) Rank: 27/87 |
0.51935 (±0.00000) Rank: 20/87 |
Chicago (CHI) | 2048.0 | — | 239.2 | 0.21284 (±0.00000) Rank: 9/87 |
0.34515 (±0.00000) Rank: 12/87 |
Helsinki (HEL) | 2048.0 | — | 526.5 | 0.39072 (±0.00000) Rank: 11/87 |
0.59304 (±0.00000) Rank: 4/87 |
Madrid (MAD) | 2048.0 | — | 209.2 | 0.07781 (±0.00000) Rank: 30/87 |
0.17751 (±0.00000) Rank: 35/87 |
Mountain View (MTV) | 2048.0 | — | 259.8 | 0.23203 (±0.00000) Rank: 19/87 |
0.37891 (±0.00000) Rank: 23/87 |
New Orleans (NOR) | 2039.2 | — | 232.5 | 0.28580 (±0.00000) Rank: 14/87 |
0.38817 (±0.00000) Rank: 20/87 |
San Francisco (SF) | 2048.0 | — | 331.3 | 0.18478 (±0.00000) Rank: 33/87 |
0.33832 (±0.00000) Rank: 31/87 |
Singapore (SG) | 2048.0 | — | 270.1 | 0.33165 (±0.00000) Rank: 19/87 |
0.45601 (±0.00000) Rank: 23/87 |
Sydney (SYD) | 2048.0 | — | 311.6 | 0.35197 (±0.00000) Rank: 11/87 |
0.51092 (±0.00000) Rank: 14/87 |
Tokyo (TOK) | 2048.0 | — | 365.8 | 0.59804 (±0.00000) Rank: 22/87 |
0.74020 (±0.00000) Rank: 22/87 |
Toronto (TOR) | 2048.0 | — | 230.4 | 0.23760 (±0.00000) Rank: 8/87 |
0.39520 (±0.00000) Rank: 7/87 |
Zurich (ZRH) | 2048.0 | — | 339.8 | 0.27835 (±0.00000) Rank: 6/87 |
0.37698 (±0.00000) Rank: 10/87 |
Average (Avg) | 2047.4 | — | 307.6 | 0.26749 (±0.00000) Rank: 14/87 |
0.40347 (±0.00000) Rank: 17/87 |
We show the inliers that survive the robust estimation loop (i.e. RANSAC), or those supplied with the submission if using custom matches. As the dataset does not contain depth, we color them based on the symmetric epopilar distance on normalized coordinates: from 0 (green) to 2e-4 (yellow), and in red above it. Note that these occasionally contain false positives.
— Amsterdam —
— Bangkok —
— Barcelona —
— Buenos Aires —
— Cambridge —
— Cannes —
— Chicago —
— Helsinki —
— Madrid —
— Mountain View —
— New Orleans —
— San Francisco —
— Singapore —
— Sydney —
— Tokyo —
— Toronto —
— Zurich —
Google Urban dataset / Multiview track
mAA at 10 degrees: N/A (±0.00000 over 1 run(s) / ±0.15698 over 17 scenes)
Rank (per category): 26 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
AMS | 2048.0 | 189.04 | 93.38 Rank: 10/87 |
1401.42 Rank: 22/87 |
3.877 Rank: 7/87 |
25.94654 Rank: 40/87 |
0.32709 (±0.00000) Rank: 35/87 |
0.43830 (±0.00000) Rank: 35/87 |
BGK | 2048.0 | 284.97 | 97.07 Rank: 14/87 |
2492.18 Rank: 16/87 |
3.683 Rank: 16/87 |
14.00159 Rank: 36/87 |
0.02993 (±0.00000) Rank: 18/87 |
0.09058 (±0.00000) Rank: 22/87 |
BCN | 2046.3 | 176.47 | 84.20 Rank: 28/87 |
1139.94 Rank: 28/87 |
3.772 Rank: 20/87 |
23.48940 Rank: 14/87 |
0.12495 (±0.00000) Rank: 30/87 |
0.19474 (±0.00000) Rank: 33/87 |
BAR | 2048.0 | 164.53 | 84.66 Rank: 30/87 |
1184.21 Rank: 29/87 |
3.342 Rank: 33/87 |
15.17802 Rank: 19/87 |
0.14608 (±0.00000) Rank: 32/87 |
0.24626 (±0.00000) Rank: 34/87 |
CAM | 2047.6 | 182.52 | 88.07 Rank: 22/87 |
1252.57 Rank: 25/87 |
3.575 Rank: 14/87 |
26.98150 Rank: 30/87 |
0.17178 (±0.00000) Rank: 26/87 |
0.26079 (±0.00000) Rank: 27/87 |
CAN | 2048.0 | 195.31 | 75.69 Rank: 31/87 |
886.38 Rank: 21/87 |
3.463 Rank: 29/87 |
21.73180 Rank: 38/87 |
0.05730 (±0.00000) Rank: 32/87 |
0.11474 (±0.00000) Rank: 29/87 |
CHI | 2048.0 | 212.97 | 88.93 Rank: 23/87 |
639.37 Rank: 18/87 |
5.038 Rank: 7/87 |
48.37715 Rank: 42/87 |
0.16746 (±0.00000) Rank: 6/87 |
0.27001 (±0.00000) Rank: 6/87 |
HEL | 2048.0 | 373.74 | 93.87 Rank: 28/87 |
2334.71 Rank: 18/87 |
4.095 Rank: 6/87 |
16.04514 Rank: 51/87 |
0.37079 (±0.00000) Rank: 6/87 |
0.50229 (±0.00000) Rank: 8/87 |
MAD | 2048.0 | 135.13 | 90.26 Rank: 27/87 |
1419.79 Rank: 19/87 |
3.320 Rank: 15/87 |
5.92477 Rank: 28/87 |
0.09325 (±0.00000) Rank: 26/87 |
0.19877 (±0.00000) Rank: 28/87 |
MTV | 2048.0 | 190.67 | 91.95 Rank: 21/87 |
1434.19 Rank: 26/87 |
3.866 Rank: 11/87 |
17.28329 Rank: 43/87 |
0.27539 (±0.00000) Rank: 24/87 |
0.40817 (±0.00000) Rank: 24/87 |
NOR | 2039.2 | 151.23 | 89.24 Rank: 12/87 |
1308.97 Rank: 11/87 |
3.432 Rank: 15/87 |
14.06611 Rank: 44/87 |
0.26941 (±0.00000) Rank: 24/87 |
0.36269 (±0.00000) Rank: 25/87 |
SF | 2048.0 | 218.67 | 96.67 Rank: 7/87 |
1758.36 Rank: 19/87 |
3.740 Rank: 22/87 |
21.88759 Rank: 48/87 |
0.28279 (±0.00000) Rank: 23/87 |
0.42897 (±0.00000) Rank: 22/87 |
SG | 2048.0 | 168.81 | 79.20 Rank: 14/87 |
1158.77 Rank: 20/87 |
3.402 Rank: 15/87 |
16.92246 Rank: 26/87 |
0.25698 (±0.00000) Rank: 7/87 |
0.33018 (±0.00000) Rank: 7/87 |
SYD | 2048.0 | 192.18 | 87.01 Rank: 31/87 |
1419.24 Rank: 27/87 |
3.575 Rank: 23/87 |
19.43343 Rank: 20/87 |
0.29596 (±0.00000) Rank: 21/87 |
0.38366 (±0.00000) Rank: 24/87 |
TOK | 2048.0 | 229.54 | 93.73 Rank: 6/87 |
2046.49 Rank: 21/87 |
3.803 Rank: 8/87 |
14.69952 Rank: 40/87 |
0.67284 (±0.00000) Rank: 15/87 |
0.75301 (±0.00000) Rank: 5/87 |
TOR | 2048.0 | 145.07 | 89.07 Rank: 14/87 |
1433.74 Rank: 22/87 |
3.255 Rank: 11/87 |
16.46908 Rank: 19/87 |
0.20179 (±0.00000) Rank: 18/87 |
0.33331 (±0.00000) Rank: 17/87 |
ZRH | 2048.0 | 209.99 | 90.98 Rank: 8/87 |
1824.98 Rank: 12/87 |
3.454 Rank: 23/87 |
21.03707 Rank: 37/87 |
0.10980 (±0.00000) Rank: 23/87 |
0.17180 (±0.00000) Rank: 29/87 |
Avg | 2047.4 | 201.23 | 89.06 Rank: 25/87 |
1478.55 Rank: 22/87 |
3.688 Rank: 11/87 |
19.96909 Rank: 25/87 |
0.22668 (±0.00000) Rank: 24/87 |
0.32284 (±0.00000) Rank: 26/87 |
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 10 images. Keypoints are drawn in blue if they are part of the model, and in red otherwise.
— Amsterdam —
— Bangkok —
— Barcelona —
— Buenos Aires —
— Cambridge —
— Cannes —
— Chicago —
— Helsinki —
— Madrid —
— Mountain View —
— New Orleans —
— San Francisco —
— Singapore —
— Sydney —
— Tokyo —
— Toronto —
— Zurich —