Submission ID: 31b7d53f
sp_scale_adapt_f_orien
Processed: 21-06-15. Download link: 31b7d53f506e274e-613json.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: sp
- Descriptor: sp (0 none: 0 bytes)
- Number of features: 2048
- Summary: sp_scale_adapt_f_orien
- Paper: N/A
- Website: N/A
- Processing date: 21-06-15
Phototourism dataset / Stereo track
mAA at 10 degrees: 0.59211 (±0.00000 over 1 run(s) / ±0.12966 over 9 scenes)
Rank (per category): 8 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
BM | 2046.0 | — | 414.4 | 0.470 Rank: 35/87 |
0.897 Rank: 39/87 |
0.35045 (±0.00000) Rank: 13/87 |
0.50411 (±0.00000) Rank: 16/87 |
FCS | 2046.8 | — | 542.3 | 0.418 Rank: 10/87 |
0.873 Rank: 17/87 |
0.65755 (±0.00000) Rank: 15/87 |
0.77509 (±0.00000) Rank: 13/87 |
LMS | 2048.0 | — | 554.9 | 0.449 Rank: 20/87 |
0.689 Rank: 35/87 |
0.62714 (±0.00000) Rank: 9/87 |
0.74666 (±0.00000) Rank: 11/87 |
LB | 2048.0 | — | 418.2 | 0.406 Rank: 20/87 |
0.704 Rank: 42/87 |
0.50107 (±0.00000) Rank: 13/87 |
0.61477 (±0.00000) Rank: 13/87 |
MC | 2048.0 | — | 453.0 | 0.429 Rank: 28/87 |
0.859 Rank: 44/87 |
0.35260 (±0.00000) Rank: 13/87 |
0.51161 (±0.00000) Rank: 20/87 |
MR | 2045.2 | — | 464.0 | 0.427 Rank: 33/87 |
0.901 Rank: 30/87 |
0.28581 (±0.00000) Rank: 13/87 |
0.40453 (±0.00000) Rank: 6/87 |
PSM | 2048.0 | — | 417.7 | 0.339 Rank: 38/87 |
0.618 Rank: 22/87 |
0.25916 (±0.00000) Rank: 11/87 |
0.42015 (±0.00000) Rank: 11/87 |
SF | 2048.0 | — | 521.1 | 0.377 Rank: 25/87 |
0.790 Rank: 32/87 |
0.51904 (±0.00000) Rank: 9/87 |
0.66282 (±0.00000) Rank: 9/87 |
SPC | 2048.0 | — | 433.4 | 0.386 Rank: 37/87 |
0.801 Rank: 32/87 |
0.52343 (±0.00000) Rank: 18/87 |
0.68922 (±0.00000) Rank: 10/87 |
Avg | 2047.3 | — | 468.8 | 0.411 Rank: 33/87 |
0.792 Rank: 26/87 |
0.45292 (±0.00000) Rank: 9/87 |
0.59211 (±0.00000) Rank: 8/87 |
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) / ±0.10156 over 9 scenes)
Rank (per category): 19 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
BM | 2046.0 | 414.98 | 100.00 Rank: 1/87 |
2094.91 Rank: 29/87 |
5.021 Rank: 23/87 |
0.34509 Rank: 18/87 |
0.60061 (±0.00000) Rank: 5/87 |
0.73904 (±0.00000) Rank: 1/87 |
FCS | 2046.8 | 549.72 | 98.27 Rank: 21/87 |
2686.49 Rank: 22/87 |
5.127 Rank: 17/87 |
0.23001 Rank: 14/87 |
0.74928 (±0.00000) Rank: 21/87 |
0.80521 (±0.00000) Rank: 25/87 |
LMS | 2048.0 | 580.48 | 99.47 Rank: 16/87 |
2020.37 Rank: 26/87 |
5.832 Rank: 24/87 |
0.29883 Rank: 45/87 |
0.87442 (±0.00000) Rank: 5/87 |
0.91942 (±0.00000) Rank: 5/87 |
LB | 2048.0 | 489.63 | 98.36 Rank: 1/87 |
2130.79 Rank: 18/87 |
5.388 Rank: 26/87 |
0.48852 Rank: 40/87 |
0.72764 (±0.00000) Rank: 10/87 |
0.81709 (±0.00000) Rank: 12/87 |
MC | 2048.0 | 445.49 | 100.00 Rank: 1/87 |
2221.51 Rank: 30/87 |
5.250 Rank: 20/87 |
0.36840 Rank: 43/87 |
0.55236 (±0.00000) Rank: 32/87 |
0.70390 (±0.00000) Rank: 22/87 |
MR | 2045.2 | 457.26 | 94.41 Rank: 38/87 |
2251.96 Rank: 34/87 |
4.883 Rank: 29/87 |
0.49030 Rank: 14/87 |
0.43771 (±0.00000) Rank: 23/87 |
0.55805 (±0.00000) Rank: 25/87 |
PSM | 2048.0 | 418.48 | 99.49 Rank: 17/87 |
2555.12 Rank: 25/87 |
4.589 Rank: 16/87 |
0.35212 Rank: 29/87 |
0.62826 (±0.00000) Rank: 39/87 |
0.71154 (±0.00000) Rank: 40/87 |
SF | 2048.0 | 513.99 | 99.68 Rank: 37/87 |
2819.05 Rank: 25/87 |
5.084 Rank: 17/87 |
0.27971 Rank: 23/87 |
0.77535 (±0.00000) Rank: 31/87 |
0.85622 (±0.00000) Rank: 34/87 |
SPC | 2048.0 | 443.38 | 100.00 Rank: 1/87 |
2323.51 Rank: 29/87 |
5.212 Rank: 20/87 |
0.41515 Rank: 13/87 |
0.75395 (±0.00000) Rank: 26/87 |
0.84769 (±0.00000) Rank: 28/87 |
Avg | 2047.3 | 479.27 | 98.85 Rank: 21/87 |
2344.86 Rank: 28/87 |
5.154 Rank: 22/87 |
0.36312 Rank: 19/87 |
0.67773 (±0.00000) Rank: 17/87 |
0.77313 (±0.00000) Rank: 19/87 |
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: 0.77654 (±0.00000 over 1 run(s) / ±0.04527 over 3 scenes)
Rank (per category): 3 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
Lizard | 2048.0 | — | 404.9 | 0.035 Rank: 32/87 |
0.012 Rank: 19/87 |
0.62524 (±0.00000) Rank: 3/87 |
0.75811 (±0.00000) Rank: 3/87 |
Pond | 2048.0 | — | 422.2 | 0.048 Rank: 41/87 |
0.046 Rank: 51/87 |
0.59760 (±0.00000) Rank: 7/87 |
0.73267 (±0.00000) Rank: 8/87 |
Tree | 2048.0 | — | 276.2 | 0.029 Rank: 39/87 |
0.024 Rank: 21/87 |
0.74614 (±0.00000) Rank: 4/87 |
0.83885 (±0.00000) Rank: 9/87 |
Avg | 2048.0 | — | 367.8 | 0.038 Rank: 36/87 |
0.027 Rank: 26/87 |
0.65632 (±0.00000) Rank: 3/87 |
0.77654 (±0.00000) Rank: 3/87 |
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) / ±0.14747 over 3 scenes)
Rank (per category): 25 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
Lizard | 2048.0 | 424.58 | 85.70 Rank: 2/87 |
1284.37 Rank: 10/87 |
3.363 Rank: 45/87 |
14.51684 Rank: 58/87 |
0.59667 (±0.00000) Rank: 8/87 |
0.67781 (±0.00000) Rank: 6/87 |
Pond | 2048.0 | 455.22 | 68.37 Rank: 34/87 |
1087.42 Rank: 7/87 |
2.958 Rank: 28/87 |
0.43981 Rank: 57/87 |
0.29042 (±0.00000) Rank: 21/87 |
0.31840 (±0.00000) Rank: 25/87 |
Tree | 2048.0 | 267.94 | 70.24 Rank: 48/87 |
571.92 Rank: 23/87 |
2.950 Rank: 47/87 |
5.87717 Rank: 21/87 |
0.43371 (±0.00000) Rank: 31/87 |
0.46674 (±0.00000) Rank: 35/87 |
Avg | 2048.0 | 382.58 | 74.77 Rank: 39/87 |
981.24 Rank: 13/87 |
3.090 Rank: 29/87 |
6.94460 Rank: 51/87 |
0.44027 (±0.00000) Rank: 19/87 |
0.48765 (±0.00000) Rank: 25/87 |
Google Urban dataset / Stereo track
mAA at 10 degrees: 0.40769 (±0.00000 over 1 run(s) / ±0.13719 over 17 scenes)
Rank (per category): 10 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
mAA(5o) | mAA(10o) |
Amsterdam (AMS) | 2048.0 | — | 287.9 | 0.27768 (±0.00000) Rank: 6/87 |
0.44420 (±0.00000) Rank: 5/87 |
Bangkok (BGK) | 2048.0 | — | 411.5 | 0.06452 (±0.00000) Rank: 15/87 |
0.18288 (±0.00000) Rank: 12/87 |
Barcelona (BCN) | 2046.3 | — | 294.8 | 0.11575 (±0.00000) Rank: 25/87 |
0.20733 (±0.00000) Rank: 24/87 |
Buenos Aires (BAR) | 2048.0 | — | 285.0 | 0.21461 (±0.00000) Rank: 6/87 |
0.35662 (±0.00000) Rank: 33/87 |
Cambridge (CAM) | 2047.6 | — | 294.5 | 0.30753 (±0.00000) Rank: 6/87 |
0.45616 (±0.00000) Rank: 23/87 |
Cannes (CAN) | 2048.0 | — | 352.8 | 0.40276 (±0.00000) Rank: 18/87 |
0.52488 (±0.00000) Rank: 16/87 |
Chicago (CHI) | 2048.0 | — | 240.2 | 0.20710 (±0.00000) Rank: 14/87 |
0.34276 (±0.00000) Rank: 14/87 |
Helsinki (HEL) | 2048.0 | — | 527.0 | 0.38814 (±0.00000) Rank: 13/87 |
0.58814 (±0.00000) Rank: 10/87 |
Madrid (MAD) | 2048.0 | — | 214.3 | 0.10091 (±0.00000) Rank: 4/87 |
0.21489 (±0.00000) Rank: 7/87 |
Mountain View (MTV) | 2048.0 | — | 262.7 | 0.22500 (±0.00000) Rank: 25/87 |
0.37539 (±0.00000) Rank: 27/87 |
New Orleans (NOR) | 2039.2 | — | 243.3 | 0.28935 (±0.00000) Rank: 7/87 |
0.41006 (±0.00000) Rank: 4/87 |
San Francisco (SF) | 2048.0 | — | 334.5 | 0.19580 (±0.00000) Rank: 23/87 |
0.33963 (±0.00000) Rank: 28/87 |
Singapore (SG) | 2048.0 | — | 279.3 | 0.32975 (±0.00000) Rank: 20/87 |
0.46076 (±0.00000) Rank: 20/87 |
Sydney (SYD) | 2048.0 | — | 316.0 | 0.36070 (±0.00000) Rank: 4/87 |
0.51572 (±0.00000) Rank: 5/87 |
Tokyo (TOK) | 2048.0 | — | 367.5 | 0.59608 (±0.00000) Rank: 23/87 |
0.74510 (±0.00000) Rank: 13/87 |
Toronto (TOR) | 2048.0 | — | 234.1 | 0.23040 (±0.00000) Rank: 12/87 |
0.38880 (±0.00000) Rank: 15/87 |
Zurich (ZRH) | 2048.0 | — | 349.1 | 0.27835 (±0.00000) Rank: 6/87 |
0.37732 (±0.00000) Rank: 7/87 |
Average (Avg) | 2047.4 | — | 311.4 | 0.26967 (±0.00000) Rank: 11/87 |
0.40769 (±0.00000) Rank: 10/87 |
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) / ±0.15040 over 17 scenes)
Rank (per category): 20 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
AMS | 2048.0 | 195.50 | 92.67 Rank: 12/87 |
1405.77 Rank: 21/87 |
3.788 Rank: 25/87 |
22.81222 Rank: 12/87 |
0.34124 (±0.00000) Rank: 30/87 |
0.45910 (±0.00000) Rank: 28/87 |
BGK | 2048.0 | 290.91 | 96.53 Rank: 26/87 |
2511.18 Rank: 14/87 |
3.641 Rank: 30/87 |
13.71305 Rank: 32/87 |
0.02942 (±0.00000) Rank: 21/87 |
0.09019 (±0.00000) Rank: 23/87 |
BCN | 2046.3 | 185.68 | 88.13 Rank: 3/87 |
1294.17 Rank: 19/87 |
3.754 Rank: 24/87 |
28.77132 Rank: 38/87 |
0.12194 (±0.00000) Rank: 32/87 |
0.20411 (±0.00000) Rank: 31/87 |
BAR | 2048.0 | 173.55 | 90.81 Rank: 1/87 |
1394.38 Rank: 3/87 |
3.388 Rank: 23/87 |
17.14594 Rank: 46/87 |
0.17427 (±0.00000) Rank: 10/87 |
0.29123 (±0.00000) Rank: 4/87 |
CAM | 2047.6 | 194.79 | 88.71 Rank: 18/87 |
1092.45 Rank: 32/87 |
3.599 Rank: 11/87 |
29.21152 Rank: 47/87 |
0.15094 (±0.00000) Rank: 32/87 |
0.22583 (±0.00000) Rank: 36/87 |
CAN | 2048.0 | 203.60 | 81.03 Rank: 12/87 |
1051.26 Rank: 6/87 |
3.508 Rank: 25/87 |
21.79995 Rank: 39/87 |
0.07142 (±0.00000) Rank: 25/87 |
0.14178 (±0.00000) Rank: 17/87 |
CHI | 2048.0 | 214.87 | 89.73 Rank: 16/87 |
640.81 Rank: 17/87 |
5.131 Rank: 5/87 |
48.11938 Rank: 38/87 |
0.16961 (±0.00000) Rank: 5/87 |
0.27667 (±0.00000) Rank: 4/87 |
HEL | 2048.0 | 382.03 | 93.33 Rank: 37/87 |
2274.25 Rank: 25/87 |
4.029 Rank: 22/87 |
15.10515 Rank: 36/87 |
0.31930 (±0.00000) Rank: 33/87 |
0.44478 (±0.00000) Rank: 33/87 |
MAD | 2048.0 | 147.66 | 92.19 Rank: 5/87 |
1528.30 Rank: 5/87 |
3.282 Rank: 26/87 |
7.68846 Rank: 48/87 |
0.10794 (±0.00000) Rank: 9/87 |
0.22039 (±0.00000) Rank: 11/87 |
MTV | 2048.0 | 198.34 | 92.67 Rank: 15/87 |
1519.15 Rank: 16/87 |
3.843 Rank: 15/87 |
15.69391 Rank: 16/87 |
0.28373 (±0.00000) Rank: 20/87 |
0.42446 (±0.00000) Rank: 16/87 |
NOR | 2039.2 | 164.56 | 87.73 Rank: 23/87 |
1249.28 Rank: 18/87 |
3.394 Rank: 20/87 |
14.17827 Rank: 46/87 |
0.26499 (±0.00000) Rank: 25/87 |
0.36405 (±0.00000) Rank: 24/87 |
SF | 2048.0 | 229.84 | 96.27 Rank: 16/87 |
1748.23 Rank: 20/87 |
3.687 Rank: 30/87 |
19.07651 Rank: 20/87 |
0.28663 (±0.00000) Rank: 21/87 |
0.42703 (±0.00000) Rank: 24/87 |
SG | 2048.0 | 179.89 | 78.04 Rank: 21/87 |
1144.75 Rank: 21/87 |
3.266 Rank: 32/87 |
16.91887 Rank: 25/87 |
0.22755 (±0.00000) Rank: 21/87 |
0.30111 (±0.00000) Rank: 21/87 |
SYD | 2048.0 | 202.17 | 90.72 Rank: 12/87 |
1586.77 Rank: 15/87 |
3.643 Rank: 8/87 |
21.84820 Rank: 37/87 |
0.32827 (±0.00000) Rank: 11/87 |
0.42741 (±0.00000) Rank: 10/87 |
TOK | 2048.0 | 235.75 | 93.73 Rank: 6/87 |
2111.38 Rank: 13/87 |
3.761 Rank: 20/87 |
14.63824 Rank: 34/87 |
0.66681 (±0.00000) Rank: 21/87 |
0.74476 (±0.00000) Rank: 16/87 |
TOR | 2048.0 | 154.49 | 91.33 Rank: 1/87 |
1467.57 Rank: 15/87 |
3.180 Rank: 36/87 |
17.28891 Rank: 31/87 |
0.17789 (±0.00000) Rank: 32/87 |
0.30487 (±0.00000) Rank: 32/87 |
ZRH | 2048.0 | 224.02 | 90.78 Rank: 12/87 |
1894.92 Rank: 5/87 |
3.441 Rank: 27/87 |
22.02223 Rank: 43/87 |
0.12710 (±0.00000) Rank: 14/87 |
0.20183 (±0.00000) Rank: 14/87 |
Avg | 2047.4 | 210.45 | 90.26 Rank: 9/87 |
1524.39 Rank: 18/87 |
3.667 Rank: 17/87 |
20.35483 Rank: 37/87 |
0.22642 (±0.00000) Rank: 25/87 |
0.32645 (±0.00000) Rank: 20/87 |
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 —









