Submission ID: d03a05f1
DISK-depth
Processed: 21-06-16. Download link: d03a05f18aef66bf-disk-depth-8000.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: Michal Tyszkiewicz (contact)
- Keypoint: disk
- Descriptor: disk (128 float32: 512 bytes)
- Number of features: 8000
- Summary: Official DISK submission with 8k features, using ratio test and cycle-consistent matching, with DEGENSAC.
- Paper: https://proceedings.neurips.cc/paper/2020/file/a42a596fc71e17828440030074d15e74-Paper.pdf
- Website: https://github.com/cvlab-epfl/disk
- Processing date: 21-06-16
Phototourism dataset / Stereo track
mAA at 10 degrees: 0.55836 (±0.00000 over 1 run(s) / ±0.12318 over 9 scenes)
Rank (per category): 17 (of 51)
| Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| BM | 7839.7 | 1990.9 | 1530.5 | 0.726 Rank: 4/51 |
0.956 Rank: 5/51 |
0.35123 (±0.00000) Rank: 9/51 |
0.49720 (±0.00000) Rank: 9/51 |
| FCS | 7996.8 | 1864.7 | 1416.1 | 0.598 Rank: 1/51 |
0.893 Rank: 26/51 |
0.61618 (±0.00000) Rank: 27/51 |
0.72579 (±0.00000) Rank: 27/51 |
| LMS | 7596.9 | 948.2 | 649.0 | 0.584 Rank: 8/51 |
0.695 Rank: 16/51 |
0.51702 (±0.00000) Rank: 28/51 |
0.62422 (±0.00000) Rank: 28/51 |
| LB | 7433.2 | 1072.2 | 809.3 | 0.662 Rank: 10/51 |
0.781 Rank: 6/51 |
0.52743 (±0.00000) Rank: 9/51 |
0.62395 (±0.00000) Rank: 16/51 |
| MC | 7887.4 | 2165.3 | 1703.4 | 0.749 Rank: 4/51 |
0.960 Rank: 6/51 |
0.42561 (±0.00000) Rank: 9/51 |
0.58186 (±0.00000) Rank: 10/51 |
| MR | 7976.1 | 1996.3 | 1612.8 | 0.676 Rank: 1/51 |
0.956 Rank: 7/51 |
0.33251 (±0.00000) Rank: 15/51 |
0.44417 (±0.00000) Rank: 15/51 |
| PSM | 7999.0 | 1141.2 | 871.1 | 0.565 Rank: 4/51 |
0.732 Rank: 10/51 |
0.16412 (±0.00000) Rank: 28/51 |
0.28545 (±0.00000) Rank: 28/51 |
| SF | 7982.0 | 1870.3 | 1407.9 | 0.592 Rank: 2/51 |
0.841 Rank: 23/51 |
0.49051 (±0.00000) Rank: 28/51 |
0.61238 (±0.00000) Rank: 30/51 |
| SPC | 7897.1 | 1546.7 | 1144.1 | 0.646 Rank: 4/51 |
0.857 Rank: 16/51 |
0.48494 (±0.00000) Rank: 18/51 |
0.63023 (±0.00000) Rank: 20/51 |
| Avg | 7845.4 | 1621.8 | 1238.2 | 0.644 Rank: 3/51 |
0.852 Rank: 8/51 |
0.43439 (±0.00000) Rank: 17/51 |
0.55836 (±0.00000) Rank: 17/51 |
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) / ±0.09245 over 9 scenes)
Rank (per category): 19 (of 51)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| BM | 7839.7 | 2002.23 | 100.00 Rank: 1/51 |
6674.58 Rank: 10/51 |
6.647 Rank: 3/51 |
0.36896 Rank: 8/51 |
0.60176 (±0.00000) Rank: 4/51 |
0.73650 (±0.00000) Rank: 4/51 |
| FCS | 7996.8 | 1927.29 | 97.53 Rank: 28/51 |
8545.17 Rank: 9/51 |
5.829 Rank: 1/51 |
0.28229 Rank: 32/51 |
0.71855 (±0.00000) Rank: 24/51 |
0.77682 (±0.00000) Rank: 27/51 |
| LMS | 7596.9 | 1029.74 | 99.67 Rank: 7/51 |
5982.19 Rank: 9/51 |
5.342 Rank: 9/51 |
0.33580 Rank: 33/51 |
0.81000 (±0.00000) Rank: 23/51 |
0.86945 (±0.00000) Rank: 24/51 |
| LB | 7433.2 | 1306.50 | 98.09 Rank: 21/51 |
5312.46 Rank: 12/51 |
6.150 Rank: 3/51 |
0.46546 Rank: 8/51 |
0.73907 (±0.00000) Rank: 9/51 |
0.81877 (±0.00000) Rank: 10/51 |
| MC | 7887.4 | 2135.04 | 100.00 Rank: 1/51 |
7393.09 Rank: 8/51 |
6.736 Rank: 2/51 |
0.35562 Rank: 22/51 |
0.54815 (±0.00000) Rank: 21/51 |
0.69506 (±0.00000) Rank: 17/51 |
| MR | 7976.1 | 1961.54 | 96.12 Rank: 6/51 |
8266.10 Rank: 8/51 |
5.914 Rank: 2/51 |
0.45865 Rank: 7/51 |
0.50053 (±0.00000) Rank: 3/51 |
0.62053 (±0.00000) Rank: 2/51 |
| PSM | 7999.0 | 1136.85 | 98.20 Rank: 23/51 |
8712.85 Rank: 9/51 |
4.499 Rank: 10/51 |
0.53835 Rank: 29/51 |
0.49987 (±0.00000) Rank: 23/51 |
0.57290 (±0.00000) Rank: 24/51 |
| SF | 7982.0 | 1841.08 | 97.51 Rank: 27/51 |
9122.96 Rank: 9/51 |
5.731 Rank: 3/51 |
0.31762 Rank: 29/51 |
0.73489 (±0.00000) Rank: 23/51 |
0.81175 (±0.00000) Rank: 25/51 |
| SPC | 7897.1 | 1606.55 | 99.48 Rank: 25/51 |
7400.64 Rank: 10/51 |
6.038 Rank: 6/51 |
0.53311 Rank: 35/51 |
0.69898 (±0.00000) Rank: 25/51 |
0.79830 (±0.00000) Rank: 26/51 |
| Avg | 7845.4 | 1660.76 | 98.51 Rank: 23/51 |
7490.01 Rank: 11/51 |
5.876 Rank: 4/51 |
0.40621 Rank: 23/51 |
0.65020 (±0.00000) Rank: 18/51 |
0.74445 (±0.00000) Rank: 19/51 |
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: 0.56223 (±0.00000 over 1 run(s) / ±0.07487 over 3 scenes)
Rank (per category): 27 (of 51)
| Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| Lizard | 8000.0 | 1067.5 | 879.5 | 0.125 Rank: 15/51 |
0.009 Rank: 28/51 |
0.36255 (±0.00000) Rank: 29/51 |
0.47531 (±0.00000) Rank: 28/51 |
| Pond | 8000.0 | 1314.0 | 1033.5 | 0.156 Rank: 15/51 |
0.055 Rank: 14/51 |
0.43006 (±0.00000) Rank: 28/51 |
0.55331 (±0.00000) Rank: 28/51 |
| Tree | 8000.0 | 856.3 | 667.1 | 0.136 Rank: 12/51 |
0.021 Rank: 13/51 |
0.54658 (±0.00000) Rank: 24/51 |
0.65806 (±0.00000) Rank: 24/51 |
| Avg | 8000.0 | 1079.2 | 860.0 | 0.139 Rank: 15/51 |
0.028 Rank: 11/51 |
0.44640 (±0.00000) Rank: 27/51 |
0.56223 (±0.00000) Rank: 27/51 |
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) / ±0.13555 over 3 scenes)
Rank (per category): 25 (of 51)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| Lizard | 8000.0 | 1327.97 | 83.50 Rank: 25/51 |
3922.59 Rank: 6/51 |
3.622 Rank: 1/51 |
13.83299 Rank: 12/51 |
0.52902 (±0.00000) Rank: 22/51 |
0.61521 (±0.00000) Rank: 23/51 |
| Pond | 8000.0 | 1919.14 | 71.60 Rank: 7/51 |
4635.21 Rank: 3/51 |
3.258 Rank: 3/51 |
0.43318 Rank: 23/51 |
0.25273 (±0.00000) Rank: 26/51 |
0.28468 (±0.00000) Rank: 28/51 |
| Tree | 8000.0 | 1137.54 | 71.83 Rank: 16/51 |
2651.22 Rank: 5/51 |
3.100 Rank: 7/51 |
6.02289 Rank: 22/51 |
0.43369 (±0.00000) Rank: 24/51 |
0.47711 (±0.00000) Rank: 20/51 |
| Avg | 8000.0 | 1461.55 | 75.64 Rank: 10/51 |
3736.34 Rank: 3/51 |
3.327 Rank: 3/51 |
6.76302 Rank: 13/51 |
0.40515 (±0.00000) Rank: 26/51 |
0.45900 (±0.00000) Rank: 25/51 |
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.
— Lizard —
— Pond —
— Tree —
Google Urban dataset / Stereo track
mAA at 10 degrees: 0.32956 (±0.00000 over 1 run(s) / ±0.12424 over 17 scenes)
Rank (per category): 16 (of 51)
| Scene | Features | Matches (raw) |
Matches (final) |
mAA(5o) | mAA(10o) |
| Amsterdam (AMS) | 7753.7 | 685.9 | 582.1 | 0.22679 (±0.00000) Rank: 14/51 |
0.34866 (±0.00000) Rank: 17/51 |
| Bangkok (BGK) | 7332.1 | 733.8 | 597.2 | 0.07692 (±0.00000) Rank: 9/51 |
0.19280 (±0.00000) Rank: 10/51 |
| Barcelona (BCN) | 7955.9 | 471.1 | 399.7 | 0.09670 (±0.00000) Rank: 21/51 |
0.19121 (±0.00000) Rank: 17/51 |
| Buenos Aires (BAR) | 7951.9 | 621.9 | 529.9 | 0.22557 (±0.00000) Rank: 14/51 |
0.36849 (±0.00000) Rank: 12/51 |
| Cambridge (CAM) | 6837.3 | 556.2 | 473.1 | 0.26575 (±0.00000) Rank: 14/51 |
0.38253 (±0.00000) Rank: 14/51 |
| Cannes (CAN) | 7491.7 | 720.5 | 588.4 | 0.38986 (±0.00000) Rank: 16/51 |
0.49401 (±0.00000) Rank: 16/51 |
| Chicago (CHI) | 7247.1 | 230.4 | 170.1 | 0.12978 (±0.00000) Rank: 17/51 |
0.20690 (±0.00000) Rank: 17/51 |
| Helsinki (HEL) | 7791.1 | 1065.1 | 846.4 | 0.33969 (±0.00000) Rank: 19/51 |
0.53995 (±0.00000) Rank: 19/51 |
| Madrid (MAD) | 7765.7 | 363.7 | 286.0 | 0.06565 (±0.00000) Rank: 14/51 |
0.14772 (±0.00000) Rank: 17/51 |
| Mountain View (MTV) | 7699.8 | 328.9 | 260.2 | 0.14727 (±0.00000) Rank: 17/51 |
0.25352 (±0.00000) Rank: 17/51 |
| New Orleans (NOR) | 7209.0 | 389.1 | 310.1 | 0.24024 (±0.00000) Rank: 12/51 |
0.30266 (±0.00000) Rank: 14/51 |
| San Francisco (SF) | 6921.2 | 534.5 | 453.5 | 0.14803 (±0.00000) Rank: 17/51 |
0.25564 (±0.00000) Rank: 17/51 |
| Singapore (SG) | 6992.6 | 471.5 | 360.5 | 0.22405 (±0.00000) Rank: 18/51 |
0.32278 (±0.00000) Rank: 18/51 |
| Sydney (SYD) | 7635.7 | 460.9 | 348.0 | 0.27424 (±0.00000) Rank: 13/51 |
0.41092 (±0.00000) Rank: 13/51 |
| Tokyo (TOK) | 7285.8 | 574.7 | 424.4 | 0.46814 (±0.00000) Rank: 14/51 |
0.60931 (±0.00000) Rank: 13/51 |
| Toronto (TOR) | 7537.0 | 247.0 | 181.2 | 0.15760 (±0.00000) Rank: 15/51 |
0.29320 (±0.00000) Rank: 14/51 |
| Zurich (ZRH) | 7480.8 | 427.6 | 332.1 | 0.22955 (±0.00000) Rank: 18/51 |
0.28213 (±0.00000) Rank: 22/51 |
| Average (Avg) | 7464.0 | 522.5 | 420.2 | 0.21799 (±0.00000) Rank: 16/51 |
0.32956 (±0.00000) Rank: 16/51 |
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) / ±0.10050 over 17 scenes)
Rank (per category): 18 (of 51)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| AMS | 7753.7 | 401.80 | 81.09 Rank: 22/51 |
3038.33 Rank: 9/51 |
3.757 Rank: 5/51 |
22.08864 Rank: 11/51 |
0.22630 (±0.00000) Rank: 21/51 |
0.31618 (±0.00000) Rank: 19/51 |
| BGK | 7332.1 | 480.05 | 94.80 Rank: 13/51 |
4774.22 Rank: 9/51 |
3.804 Rank: 3/51 |
14.11177 Rank: 19/51 |
0.02084 (±0.00000) Rank: 20/51 |
0.07078 (±0.00000) Rank: 21/51 |
| BCN | 7955.9 | 270.14 | 83.20 Rank: 10/51 |
1767.40 Rank: 10/51 |
3.476 Rank: 9/51 |
27.56518 Rank: 19/51 |
0.09495 (±0.00000) Rank: 16/51 |
0.16375 (±0.00000) Rank: 16/51 |
| BAR | 7951.9 | 334.88 | 78.39 Rank: 12/51 |
2161.88 Rank: 9/51 |
3.099 Rank: 14/51 |
15.24081 Rank: 18/51 |
0.12473 (±0.00000) Rank: 11/51 |
0.19390 (±0.00000) Rank: 11/51 |
| CAM | 6837.3 | 319.61 | 80.78 Rank: 12/51 |
2691.43 Rank: 9/51 |
3.382 Rank: 6/51 |
25.15723 Rank: 20/51 |
0.10625 (±0.00000) Rank: 19/51 |
0.16358 (±0.00000) Rank: 19/51 |
| CAN | 7491.7 | 388.39 | 74.02 Rank: 11/51 |
1506.57 Rank: 9/51 |
3.445 Rank: 4/51 |
19.92061 Rank: 19/51 |
0.06925 (±0.00000) Rank: 10/51 |
0.10990 (±0.00000) Rank: 10/51 |
| CHI | 7247.1 | 198.12 | 86.72 Rank: 12/51 |
765.25 Rank: 9/51 |
4.899 Rank: 8/51 |
47.99906 Rank: 24/51 |
0.09974 (±0.00000) Rank: 12/51 |
0.17601 (±0.00000) Rank: 12/51 |
| HEL | 7791.1 | 757.57 | 91.67 Rank: 14/51 |
5314.34 Rank: 8/51 |
3.668 Rank: 7/51 |
11.19858 Rank: 3/51 |
0.20231 (±0.00000) Rank: 26/51 |
0.32292 (±0.00000) Rank: 26/51 |
| MAD | 7765.7 | 226.52 | 80.37 Rank: 19/51 |
1925.74 Rank: 13/51 |
3.414 Rank: 3/51 |
3.99682 Rank: 6/51 |
0.02379 (±0.00000) Rank: 25/51 |
0.06479 (±0.00000) Rank: 24/51 |
| MTV | 7699.8 | 233.33 | 83.66 Rank: 16/51 |
1808.38 Rank: 13/51 |
3.699 Rank: 10/51 |
12.85267 Rank: 7/51 |
0.11934 (±0.00000) Rank: 20/51 |
0.22182 (±0.00000) Rank: 19/51 |
| NOR | 7209.0 | 253.38 | 63.62 Rank: 17/51 |
1251.04 Rank: 11/51 |
3.032 Rank: 13/51 |
9.28906 Rank: 15/51 |
0.09947 (±0.00000) Rank: 14/51 |
0.14309 (±0.00000) Rank: 15/51 |
| SF | 6921.2 | 304.87 | 80.37 Rank: 18/51 |
2145.28 Rank: 11/51 |
3.279 Rank: 15/51 |
11.39251 Rank: 6/51 |
0.11956 (±0.00000) Rank: 19/51 |
0.19734 (±0.00000) Rank: 19/51 |
| SG | 6992.6 | 284.19 | 62.86 Rank: 20/51 |
1358.27 Rank: 12/51 |
3.229 Rank: 5/51 |
12.33158 Rank: 6/51 |
0.05427 (±0.00000) Rank: 23/51 |
0.08554 (±0.00000) Rank: 23/51 |
| SYD | 7635.7 | 262.99 | 80.98 Rank: 11/51 |
1491.29 Rank: 11/51 |
3.351 Rank: 7/51 |
16.44573 Rank: 24/51 |
0.15716 (±0.00000) Rank: 15/51 |
0.24471 (±0.00000) Rank: 13/51 |
| TOK | 7285.8 | 350.11 | 83.05 Rank: 19/51 |
3286.18 Rank: 9/51 |
3.608 Rank: 6/51 |
11.52320 Rank: 10/51 |
0.37527 (±0.00000) Rank: 21/51 |
0.43894 (±0.00000) Rank: 21/51 |
| TOR | 7537.0 | 142.09 | 78.28 Rank: 12/51 |
1054.58 Rank: 14/51 |
3.226 Rank: 8/51 |
14.96775 Rank: 18/51 |
0.08364 (±0.00000) Rank: 15/51 |
0.15645 (±0.00000) Rank: 15/51 |
| ZRH | 7480.8 | 251.63 | 65.90 Rank: 23/51 |
2005.17 Rank: 12/51 |
3.056 Rank: 13/51 |
12.52931 Rank: 5/51 |
0.01615 (±0.00000) Rank: 26/51 |
0.04611 (±0.00000) Rank: 25/51 |
| Avg | 7464.0 | 321.16 | 79.40 Rank: 13/51 |
2255.61 Rank: 9/51 |
3.496 Rank: 7/51 |
16.97709 Rank: 18/51 |
0.11724 (±0.00000) Rank: 19/51 |
0.18328 (±0.00000) Rank: 18/51 |
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 —
