Submission ID: 7f08a759
Modified-DISK-TransUnet
Processed: 21-06-10. Download link: 7f08a759d4d62ba2-disk-transunet.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: Yuhao Zhou, Xiaohu Nan, Xudong Zhang (contact)
- Keypoint: disk
- Descriptor: disk (128 float32: 512 bytes)
- Number of features: 2048
- Summary:
- Paper: N/A
- Website: N/A
- Processing date: 21-06-10
Phototourism dataset / Stereo track
mAA at 10 degrees: 0.46398 (±0.00000 over 1 run(s) / ±0.15427 over 9 scenes)
Rank (per category): 53 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
BM | 2048.0 | — | 991.3 | 0.652 Rank: 2/87 |
0.932 Rank: 18/87 |
0.20102 (±0.00000) Rank: 57/87 |
0.33178 (±0.00000) Rank: 56/87 |
FCS | 2019.1 | — | 524.9 | 0.413 Rank: 27/87 |
0.852 Rank: 59/87 |
0.54192 (±0.00000) Rank: 51/87 |
0.66524 (±0.00000) Rank: 51/87 |
LMS | 2048.0 | — | 366.8 | 0.357 Rank: 70/87 |
0.687 Rank: 38/87 |
0.49928 (±0.00000) Rank: 55/87 |
0.61659 (±0.00000) Rank: 56/87 |
LB | 2048.0 | — | 479.8 | 0.480 Rank: 1/87 |
0.714 Rank: 22/87 |
0.40022 (±0.00000) Rank: 57/87 |
0.49082 (±0.00000) Rank: 56/87 |
MC | 2048.0 | — | 786.4 | 0.541 Rank: 8/87 |
0.930 Rank: 13/87 |
0.25080 (±0.00000) Rank: 57/87 |
0.40647 (±0.00000) Rank: 59/87 |
MR | 2048.0 | — | 492.1 | 0.417 Rank: 45/87 |
0.838 Rank: 70/87 |
0.20905 (±0.00000) Rank: 59/87 |
0.28603 (±0.00000) Rank: 69/87 |
PSM | 2048.0 | — | 430.8 | 0.374 Rank: 8/87 |
0.590 Rank: 47/87 |
0.11219 (±0.00000) Rank: 60/87 |
0.20760 (±0.00000) Rank: 59/87 |
SF | 2048.0 | — | 509.0 | 0.392 Rank: 15/87 |
0.796 Rank: 26/87 |
0.42650 (±0.00000) Rank: 49/87 |
0.55504 (±0.00000) Rank: 53/87 |
SPC | 2048.0 | — | 648.5 | 0.484 Rank: 1/87 |
0.851 Rank: 7/87 |
0.46309 (±0.00000) Rank: 39/87 |
0.61629 (±0.00000) Rank: 41/87 |
Avg | 2044.8 | — | 581.1 | 0.457 Rank: 3/87 |
0.799 Rank: 19/87 |
0.34490 (±0.00000) Rank: 52/87 |
0.46398 (±0.00000) Rank: 53/87 |
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) / ±0.12425 over 9 scenes)
Rank (per category): 50 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
BM | 2048.0 | 996.95 | 99.60 Rank: 65/87 |
2584.69 Rank: 1/87 |
7.202 Rank: 2/87 |
0.45543 Rank: 52/87 |
0.44240 (±0.00000) Rank: 56/87 |
0.61346 (±0.00000) Rank: 52/87 |
FCS | 2019.1 | 543.84 | 96.89 Rank: 61/87 |
2691.74 Rank: 21/87 |
5.333 Rank: 5/87 |
0.27158 Rank: 60/87 |
0.69927 (±0.00000) Rank: 58/87 |
0.76199 (±0.00000) Rank: 62/87 |
LMS | 2048.0 | 388.44 | 98.93 Rank: 56/87 |
2049.03 Rank: 20/87 |
5.226 Rank: 59/87 |
0.30914 Rank: 53/87 |
0.80288 (±0.00000) Rank: 57/87 |
0.85719 (±0.00000) Rank: 60/87 |
LB | 2048.0 | 562.93 | 97.96 Rank: 48/87 |
2313.98 Rank: 4/87 |
5.902 Rank: 9/87 |
0.47485 Rank: 17/87 |
0.71419 (±0.00000) Rank: 32/87 |
0.79934 (±0.00000) Rank: 39/87 |
MC | 2048.0 | 769.64 | 99.93 Rank: 43/87 |
2781.86 Rank: 2/87 |
6.354 Rank: 6/87 |
0.38424 Rank: 51/87 |
0.51043 (±0.00000) Rank: 53/87 |
0.66823 (±0.00000) Rank: 47/87 |
MR | 2048.0 | 482.74 | 93.91 Rank: 47/87 |
2524.47 Rank: 9/87 |
5.054 Rank: 2/87 |
0.64516 Rank: 76/87 |
0.37336 (±0.00000) Rank: 55/87 |
0.49447 (±0.00000) Rank: 54/87 |
PSM | 2048.0 | 427.49 | 96.95 Rank: 45/87 |
3148.69 Rank: 2/87 |
4.222 Rank: 48/87 |
0.55319 Rank: 57/87 |
0.46013 (±0.00000) Rank: 54/87 |
0.54627 (±0.00000) Rank: 53/87 |
SF | 2048.0 | 501.71 | 96.85 Rank: 59/87 |
2795.42 Rank: 30/87 |
5.297 Rank: 2/87 |
0.28942 Rank: 37/87 |
0.72497 (±0.00000) Rank: 52/87 |
0.80924 (±0.00000) Rank: 53/87 |
SPC | 2048.0 | 663.42 | 99.80 Rank: 49/87 |
2725.94 Rank: 2/87 |
5.986 Rank: 4/87 |
0.44852 Rank: 39/87 |
0.72405 (±0.00000) Rank: 41/87 |
0.82157 (±0.00000) Rank: 41/87 |
Avg | 2044.8 | 593.02 | 97.87 Rank: 51/87 |
2623.98 Rank: 2/87 |
5.620 Rank: 6/87 |
0.42572 Rank: 56/87 |
0.60574 (±0.00000) Rank: 52/87 |
0.70797 (±0.00000) Rank: 50/87 |
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: 0.48612 (±0.00000 over 1 run(s) / ±0.09991 over 3 scenes)
Rank (per category): 60 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
Lizard | 2048.0 | — | 220.1 | 0.040 Rank: 18/87 |
0.004 Rank: 72/87 |
0.27961 (±0.00000) Rank: 66/87 |
0.38169 (±0.00000) Rank: 65/87 |
Pond | 2048.0 | — | 278.0 | 0.061 Rank: 16/87 |
0.055 Rank: 13/87 |
0.35631 (±0.00000) Rank: 60/87 |
0.45591 (±0.00000) Rank: 63/87 |
Tree | 2048.0 | — | 202.1 | 0.037 Rank: 17/87 |
0.016 Rank: 71/87 |
0.53024 (±0.00000) Rank: 49/87 |
0.62075 (±0.00000) Rank: 51/87 |
Avg | 2048.0 | — | 233.4 | 0.046 Rank: 16/87 |
0.025 Rank: 67/87 |
0.38872 (±0.00000) Rank: 60/87 |
0.48612 (±0.00000) Rank: 60/87 |
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) / ±0.12106 over 3 scenes)
Rank (per category): 54 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
Lizard | 2048.0 | 255.58 | 83.43 Rank: 55/87 |
832.79 Rank: 49/87 |
3.433 Rank: 14/87 |
14.08590 Rank: 26/87 |
0.51538 (±0.00000) Rank: 57/87 |
0.60756 (±0.00000) Rank: 57/87 |
Pond | 2048.0 | 367.18 | 66.43 Rank: 61/87 |
883.91 Rank: 36/87 |
2.996 Rank: 20/87 |
0.43727 Rank: 52/87 |
0.27591 (±0.00000) Rank: 41/87 |
0.31211 (±0.00000) Rank: 37/87 |
Tree | 2048.0 | 241.26 | 71.61 Rank: 26/87 |
579.80 Rank: 19/87 |
3.167 Rank: 3/87 |
6.06573 Rank: 41/87 |
0.39784 (±0.00000) Rank: 52/87 |
0.43788 (±0.00000) Rank: 51/87 |
Avg | 2048.0 | 288.01 | 73.83 Rank: 57/87 |
765.50 Rank: 44/87 |
3.198 Rank: 7/87 |
6.86297 Rank: 29/87 |
0.39638 (±0.00000) Rank: 56/87 |
0.45251 (±0.00000) Rank: 54/87 |
Google Urban dataset / Stereo track
mAA at 10 degrees: 0.27752 (±0.00000 over 1 run(s) / ±0.10574 over 17 scenes)
Rank (per category): 45 (of 87)
Scene | Features | Matches (raw) |
Matches (final) |
mAA(5o) | mAA(10o) |
Amsterdam (AMS) | 2040.7 | — | 312.3 | 0.14286 (±0.00000) Rank: 45/87 |
0.26339 (±0.00000) Rank: 44/87 |
Bangkok (BGK) | 2046.3 | — | 378.3 | 0.07246 (±0.00000) Rank: 8/87 |
0.19479 (±0.00000) Rank: 1/87 |
Barcelona (BCN) | 2028.5 | — | 224.0 | 0.06007 (±0.00000) Rank: 47/87 |
0.11758 (±0.00000) Rank: 47/87 |
Buenos Aires (BAR) | 2048.0 | — | 218.1 | 0.15342 (±0.00000) Rank: 45/87 |
0.27078 (±0.00000) Rank: 45/87 |
Cambridge (CAM) | 2033.1 | — | 251.9 | 0.18425 (±0.00000) Rank: 45/87 |
0.30685 (±0.00000) Rank: 47/87 |
Cannes (CAN) | 2000.0 | — | 285.8 | 0.29124 (±0.00000) Rank: 45/87 |
0.38203 (±0.00000) Rank: 45/87 |
Chicago (CHI) | 1995.8 | — | 152.9 | 0.08634 (±0.00000) Rank: 43/87 |
0.15000 (±0.00000) Rank: 42/87 |
Helsinki (HEL) | 2048.0 | — | 494.1 | 0.32216 (±0.00000) Rank: 38/87 |
0.51830 (±0.00000) Rank: 37/87 |
Madrid (MAD) | 2048.0 | — | 173.3 | 0.06201 (±0.00000) Rank: 41/87 |
0.16596 (±0.00000) Rank: 38/87 |
Mountain View (MTV) | 2031.7 | — | 217.7 | 0.13672 (±0.00000) Rank: 45/87 |
0.23691 (±0.00000) Rank: 44/87 |
New Orleans (NOR) | 2047.6 | — | 164.4 | 0.19053 (±0.00000) Rank: 45/87 |
0.25207 (±0.00000) Rank: 45/87 |
San Francisco (SF) | 2041.4 | — | 226.0 | 0.13648 (±0.00000) Rank: 44/87 |
0.24698 (±0.00000) Rank: 45/87 |
Singapore (SG) | 2015.7 | — | 202.2 | 0.17278 (±0.00000) Rank: 47/87 |
0.26899 (±0.00000) Rank: 47/87 |
Sydney (SYD) | 2025.0 | — | 184.3 | 0.20611 (±0.00000) Rank: 44/87 |
0.33974 (±0.00000) Rank: 43/87 |
Tokyo (TOK) | 2048.0 | — | 243.9 | 0.34804 (±0.00000) Rank: 48/87 |
0.49608 (±0.00000) Rank: 48/87 |
Toronto (TOR) | 2042.7 | — | 139.9 | 0.15680 (±0.00000) Rank: 40/87 |
0.29160 (±0.00000) Rank: 40/87 |
Zurich (ZRH) | 2034.7 | — | 230.9 | 0.15601 (±0.00000) Rank: 47/87 |
0.21581 (±0.00000) Rank: 47/87 |
Average (Avg) | 2033.8 | — | 241.2 | 0.16931 (±0.00000) Rank: 46/87 |
0.27752 (±0.00000) Rank: 45/87 |
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) / ±0.09567 over 17 scenes)
Rank (per category): 44 (of 87)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
AMS | 2040.7 | 192.29 | 84.03 Rank: 42/87 |
1488.67 Rank: 10/87 |
3.606 Rank: 42/87 |
19.89419 Rank: 8/87 |
0.20951 (±0.00000) Rank: 44/87 |
0.30131 (±0.00000) Rank: 43/87 |
BGK | 2046.3 | 255.04 | 94.67 Rank: 41/87 |
2459.35 Rank: 21/87 |
3.654 Rank: 27/87 |
13.83643 Rank: 34/87 |
0.02688 (±0.00000) Rank: 29/87 |
0.09124 (±0.00000) Rank: 20/87 |
BCN | 2028.5 | 127.96 | 77.64 Rank: 43/87 |
1041.27 Rank: 30/87 |
3.459 Rank: 45/87 |
21.95424 Rank: 6/87 |
0.07996 (±0.00000) Rank: 43/87 |
0.12499 (±0.00000) Rank: 45/87 |
BAR | 2048.0 | 123.63 | 72.78 Rank: 43/87 |
862.76 Rank: 34/87 |
3.091 Rank: 44/87 |
14.16212 Rank: 12/87 |
0.08787 (±0.00000) Rank: 42/87 |
0.16026 (±0.00000) Rank: 42/87 |
CAM | 2033.1 | 151.82 | 76.00 Rank: 43/87 |
1305.15 Rank: 17/87 |
3.033 Rank: 49/87 |
18.11527 Rank: 7/87 |
0.10047 (±0.00000) Rank: 46/87 |
0.15985 (±0.00000) Rank: 45/87 |
CAN | 2000.0 | 156.16 | 69.50 Rank: 37/87 |
692.60 Rank: 32/87 |
3.349 Rank: 38/87 |
15.04734 Rank: 13/87 |
0.02056 (±0.00000) Rank: 46/87 |
0.05441 (±0.00000) Rank: 42/87 |
CHI | 1995.8 | 133.53 | 85.21 Rank: 32/87 |
413.05 Rank: 32/87 |
4.844 Rank: 18/87 |
45.00981 Rank: 12/87 |
0.07156 (±0.00000) Rank: 34/87 |
0.13800 (±0.00000) Rank: 34/87 |
HEL | 2048.0 | 351.55 | 95.66 Rank: 4/87 |
2691.77 Rank: 1/87 |
3.651 Rank: 46/87 |
9.97354 Rank: 3/87 |
0.25616 (±0.00000) Rank: 43/87 |
0.38296 (±0.00000) Rank: 43/87 |
MAD | 2048.0 | 108.96 | 81.25 Rank: 43/87 |
1097.23 Rank: 34/87 |
3.132 Rank: 46/87 |
4.28076 Rank: 7/87 |
0.03723 (±0.00000) Rank: 44/87 |
0.08879 (±0.00000) Rank: 44/87 |
MTV | 2031.7 | 155.99 | 85.38 Rank: 42/87 |
1323.10 Rank: 31/87 |
3.771 Rank: 37/87 |
14.44044 Rank: 9/87 |
0.14684 (±0.00000) Rank: 44/87 |
0.24816 (±0.00000) Rank: 43/87 |
NOR | 2047.6 | 107.40 | 64.72 Rank: 43/87 |
747.25 Rank: 32/87 |
3.154 Rank: 41/87 |
8.03535 Rank: 10/87 |
0.11524 (±0.00000) Rank: 42/87 |
0.15495 (±0.00000) Rank: 42/87 |
SF | 2041.4 | 143.54 | 82.99 Rank: 43/87 |
1024.77 Rank: 35/87 |
3.407 Rank: 44/87 |
15.31184 Rank: 11/87 |
0.10042 (±0.00000) Rank: 48/87 |
0.16951 (±0.00000) Rank: 48/87 |
SG | 2015.7 | 126.03 | 62.64 Rank: 45/87 |
757.74 Rank: 34/87 |
3.095 Rank: 47/87 |
12.69226 Rank: 6/87 |
0.06057 (±0.00000) Rank: 47/87 |
0.09812 (±0.00000) Rank: 47/87 |
SYD | 2025.0 | 113.03 | 72.40 Rank: 43/87 |
745.75 Rank: 42/87 |
3.100 Rank: 46/87 |
12.08669 Rank: 9/87 |
0.11879 (±0.00000) Rank: 44/87 |
0.17863 (±0.00000) Rank: 45/87 |
TOK | 2048.0 | 149.96 | 77.83 Rank: 48/87 |
1445.17 Rank: 38/87 |
3.571 Rank: 44/87 |
9.68943 Rank: 5/87 |
0.30005 (±0.00000) Rank: 48/87 |
0.36997 (±0.00000) Rank: 48/87 |
TOR | 2042.7 | 86.94 | 75.73 Rank: 42/87 |
780.15 Rank: 40/87 |
2.978 Rank: 47/87 |
12.10578 Rank: 10/87 |
0.07665 (±0.00000) Rank: 42/87 |
0.14868 (±0.00000) Rank: 42/87 |
ZRH | 2034.7 | 138.23 | 70.73 Rank: 46/87 |
988.55 Rank: 40/87 |
3.048 Rank: 50/87 |
13.51797 Rank: 7/87 |
0.02915 (±0.00000) Rank: 47/87 |
0.06192 (±0.00000) Rank: 47/87 |
Avg | 2033.8 | 154.24 | 78.19 Rank: 43/87 |
1168.49 Rank: 32/87 |
3.408 Rank: 42/87 |
15.30315 Rank: 10/87 |
0.10811 (±0.00000) Rank: 46/87 |
0.17246 (±0.00000) Rank: 44/87 |
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 —









