Submission ID: a77ea312
HarrisZ improved AffNet-HardNet8
Processed: 21-06-11. Download link: a77ea3121a869484-harrisz.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: Fabio Bellavia and Dmytro Mishkin (contact)
- Keypoint: hzbnn22l05m8k2l
- Descriptor: hardnet (128 float32: 512 bytes)
- Number of features: 8000
- Summary: Baseline submission for HarrisZ detector evaluation
- Paper: https://digital-library.theiet.org/content/journals/10.1049/iet-cvi.2009.0127
- Website: N/A
- Processing date: 21-06-11
Phototourism dataset / Stereo track
mAA at 10 degrees: 0.55788 (±0.00000 over 1 run(s) / ±0.14219 over 9 scenes)
Rank (per category): 18 (of 51)
| Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| BM | 5924.1 | 577.2 | 422.5 | 0.641 Rank: 20/51 |
0.953 Rank: 6/51 |
0.27851 (±0.00000) Rank: 17/51 |
0.41083 (±0.00000) Rank: 18/51 |
| FCS | 7673.2 | 771.4 | 587.6 | 0.522 Rank: 14/51 |
0.936 Rank: 3/51 |
0.63191 (±0.00000) Rank: 23/51 |
0.74787 (±0.00000) Rank: 24/51 |
| LMS | 5167.7 | 561.6 | 394.3 | 0.519 Rank: 20/51 |
0.708 Rank: 12/51 |
0.58585 (±0.00000) Rank: 15/51 |
0.71425 (±0.00000) Rank: 11/51 |
| LB | 7137.0 | 512.1 | 350.4 | 0.531 Rank: 24/51 |
0.806 Rank: 1/51 |
0.50940 (±0.00000) Rank: 17/51 |
0.61447 (±0.00000) Rank: 19/51 |
| MC | 7683.4 | 842.5 | 648.6 | 0.627 Rank: 20/51 |
0.958 Rank: 7/51 |
0.38904 (±0.00000) Rank: 17/51 |
0.54064 (±0.00000) Rank: 27/51 |
| MR | 7601.6 | 884.7 | 674.5 | 0.536 Rank: 25/51 |
0.958 Rank: 6/51 |
0.31240 (±0.00000) Rank: 25/51 |
0.42106 (±0.00000) Rank: 26/51 |
| PSM | 7643.2 | 393.4 | 293.2 | 0.476 Rank: 20/51 |
0.791 Rank: 5/51 |
0.17742 (±0.00000) Rank: 23/51 |
0.30211 (±0.00000) Rank: 24/51 |
| SF | 7832.8 | 999.2 | 771.0 | 0.513 Rank: 22/51 |
0.887 Rank: 3/51 |
0.51834 (±0.00000) Rank: 26/51 |
0.64714 (±0.00000) Rank: 26/51 |
| SPC | 7118.8 | 590.4 | 402.9 | 0.568 Rank: 20/51 |
0.896 Rank: 3/51 |
0.47194 (±0.00000) Rank: 21/51 |
0.62255 (±0.00000) Rank: 22/51 |
| Avg | 7086.9 | 681.4 | 505.0 | 0.548 Rank: 20/51 |
0.877 Rank: 2/51 |
0.43054 (±0.00000) Rank: 18/51 |
0.55788 (±0.00000) Rank: 18/51 |
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) / ±0.12607 over 9 scenes)
Rank (per category): 21 (of 51)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| BM | 5924.1 | 853.02 | 99.81 Rank: 22/51 |
2973.07 Rank: 33/51 |
4.863 Rank: 30/51 |
0.46214 Rank: 25/51 |
0.48536 (±0.00000) Rank: 26/51 |
0.62051 (±0.00000) Rank: 29/51 |
| FCS | 7673.2 | 1189.26 | 99.27 Rank: 2/51 |
6066.17 Rank: 17/51 |
4.582 Rank: 24/51 |
0.22512 Rank: 12/51 |
0.76350 (±0.00000) Rank: 15/51 |
0.82591 (±0.00000) Rank: 10/51 |
| LMS | 5167.7 | 882.47 | 99.68 Rank: 5/51 |
3190.20 Rank: 24/51 |
4.994 Rank: 21/51 |
0.31551 Rank: 24/51 |
0.85053 (±0.00000) Rank: 9/51 |
0.90529 (±0.00000) Rank: 6/51 |
| LB | 7137.0 | 958.72 | 98.03 Rank: 24/51 |
3706.49 Rank: 25/51 |
4.711 Rank: 28/51 |
0.46146 Rank: 6/51 |
0.71803 (±0.00000) Rank: 16/51 |
0.80271 (±0.00000) Rank: 17/51 |
| MC | 7683.4 | 1201.29 | 99.81 Rank: 26/51 |
5033.70 Rank: 19/51 |
5.026 Rank: 27/51 |
0.38065 Rank: 34/51 |
0.53161 (±0.00000) Rank: 29/51 |
0.67795 (±0.00000) Rank: 31/51 |
| MR | 7601.6 | 1198.21 | 97.13 Rank: 3/51 |
5591.06 Rank: 22/51 |
4.630 Rank: 23/51 |
0.49137 Rank: 22/51 |
0.46928 (±0.00000) Rank: 16/51 |
0.59119 (±0.00000) Rank: 12/51 |
| PSM | 7643.2 | 723.00 | 98.39 Rank: 21/51 |
6301.46 Rank: 20/51 |
3.414 Rank: 29/51 |
0.55622 Rank: 30/51 |
0.44520 (±0.00000) Rank: 30/51 |
0.52968 (±0.00000) Rank: 30/51 |
| SF | 7832.8 | 1336.33 | 99.43 Rank: 18/51 |
6899.75 Rank: 17/51 |
4.686 Rank: 28/51 |
0.28317 Rank: 8/51 |
0.77150 (±0.00000) Rank: 15/51 |
0.85468 (±0.00000) Rank: 12/51 |
| SPC | 7118.8 | 965.42 | 99.87 Rank: 20/51 |
4626.02 Rank: 24/51 |
4.600 Rank: 29/51 |
0.53279 Rank: 34/51 |
0.72355 (±0.00000) Rank: 16/51 |
0.82280 (±0.00000) Rank: 17/51 |
| Avg | 7086.9 | 1034.19 | 99.05 Rank: 6/51 |
4931.99 Rank: 22/51 |
4.612 Rank: 27/51 |
0.41205 Rank: 29/51 |
0.63984 (±0.00000) Rank: 21/51 |
0.73675 (±0.00000) Rank: 21/51 |
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: 0.73683 (±0.00000 over 1 run(s) / ±0.06664 over 3 scenes)
Rank (per category): 11 (of 51)
| Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| Lizard | 8000.0 | 721.6 | 680.4 | 0.084 Rank: 34/51 |
0.010 Rank: 22/51 |
0.56449 (±0.00000) Rank: 11/51 |
0.69723 (±0.00000) Rank: 11/51 |
| Pond | 8000.0 | 691.3 | 632.0 | 0.108 Rank: 29/51 |
0.052 Rank: 23/51 |
0.55110 (±0.00000) Rank: 17/51 |
0.68257 (±0.00000) Rank: 14/51 |
| Tree | 8000.0 | 499.9 | 451.0 | 0.084 Rank: 34/51 |
0.023 Rank: 4/51 |
0.74834 (±0.00000) Rank: 12/51 |
0.83068 (±0.00000) Rank: 12/51 |
| Avg | 8000.0 | 637.6 | 587.8 | 0.092 Rank: 34/51 |
0.028 Rank: 12/51 |
0.62131 (±0.00000) Rank: 12/51 |
0.73683 (±0.00000) Rank: 11/51 |
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) / ±0.15319 over 3 scenes)
Rank (per category): 16 (of 51)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| Lizard | 8000.0 | 1126.66 | 85.30 Rank: 4/51 |
2765.29 Rank: 15/51 |
3.407 Rank: 15/51 |
14.35137 Rank: 38/51 |
0.58147 (±0.00000) Rank: 8/51 |
0.66371 (±0.00000) Rank: 8/51 |
| Pond | 8000.0 | 1396.23 | 71.23 Rank: 9/51 |
2985.77 Rank: 12/51 |
3.133 Rank: 9/51 |
0.43799 Rank: 30/51 |
0.26318 (±0.00000) Rank: 23/51 |
0.28864 (±0.00000) Rank: 24/51 |
| Tree | 8000.0 | 864.97 | 71.62 Rank: 18/51 |
1449.50 Rank: 16/51 |
2.990 Rank: 18/51 |
6.04097 Rank: 23/51 |
0.42711 (±0.00000) Rank: 27/51 |
0.46647 (±0.00000) Rank: 24/51 |
| Avg | 8000.0 | 1129.29 | 76.05 Rank: 7/51 |
2400.19 Rank: 15/51 |
3.177 Rank: 12/51 |
6.94344 Rank: 30/51 |
0.42392 (±0.00000) Rank: 17/51 |
0.47294 (±0.00000) Rank: 16/51 |
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: 0.35672 (±0.00000 over 1 run(s) / ±0.12967 over 17 scenes)
Rank (per category): 12 (of 51)
| Scene | Features | Matches (raw) |
Matches (final) |
mAA(5o) | mAA(10o) |
| Amsterdam (AMS) | 4306.7 | 276.5 | 202.4 | 0.22054 (±0.00000) Rank: 17/51 |
0.34866 (±0.00000) Rank: 17/51 |
| Bangkok (BGK) | 3087.8 | 239.9 | 165.9 | 0.04467 (±0.00000) Rank: 23/51 |
0.13499 (±0.00000) Rank: 24/51 |
| Barcelona (BCN) | 4576.9 | 243.4 | 181.4 | 0.16557 (±0.00000) Rank: 1/51 |
0.25275 (±0.00000) Rank: 1/51 |
| Buenos Aires (BAR) | 3994.0 | 272.2 | 192.5 | 0.21553 (±0.00000) Rank: 17/51 |
0.33516 (±0.00000) Rank: 14/51 |
| Cambridge (CAM) | 2981.6 | 155.8 | 109.2 | 0.30822 (±0.00000) Rank: 17/51 |
0.43116 (±0.00000) Rank: 5/51 |
| Cannes (CAN) | 3345.9 | 271.1 | 186.5 | 0.40369 (±0.00000) Rank: 13/51 |
0.50922 (±0.00000) Rank: 5/51 |
| Chicago (CHI) | 4274.9 | 129.5 | 83.6 | 0.23402 (±0.00000) Rank: 8/51 |
0.34829 (±0.00000) Rank: 9/51 |
| Helsinki (HEL) | 3348.7 | 342.4 | 229.9 | 0.39021 (±0.00000) Rank: 3/51 |
0.59098 (±0.00000) Rank: 2/51 |
| Madrid (MAD) | 3668.0 | 125.2 | 76.7 | 0.05289 (±0.00000) Rank: 21/51 |
0.12796 (±0.00000) Rank: 21/51 |
| Mountain View (MTV) | 3896.6 | 133.6 | 85.6 | 0.17891 (±0.00000) Rank: 13/51 |
0.29316 (±0.00000) Rank: 13/51 |
| New Orleans (NOR) | 3534.6 | 166.1 | 110.3 | 0.23669 (±0.00000) Rank: 13/51 |
0.31627 (±0.00000) Rank: 12/51 |
| San Francisco (SF) | 3314.3 | 234.7 | 176.9 | 0.15906 (±0.00000) Rank: 15/51 |
0.28898 (±0.00000) Rank: 14/51 |
| Singapore (SG) | 2854.6 | 211.2 | 142.1 | 0.30506 (±0.00000) Rank: 11/51 |
0.42120 (±0.00000) Rank: 5/51 |
| Sydney (SYD) | 3081.6 | 180.1 | 119.8 | 0.29782 (±0.00000) Rank: 12/51 |
0.42838 (±0.00000) Rank: 12/51 |
| Tokyo (TOK) | 3047.8 | 253.6 | 160.6 | 0.47500 (±0.00000) Rank: 12/51 |
0.61201 (±0.00000) Rank: 12/51 |
| Toronto (TOR) | 3514.5 | 115.7 | 71.8 | 0.16400 (±0.00000) Rank: 14/51 |
0.29480 (±0.00000) Rank: 13/51 |
| Zurich (ZRH) | 4168.4 | 266.7 | 191.0 | 0.25636 (±0.00000) Rank: 6/51 |
0.33024 (±0.00000) Rank: 6/51 |
| Average (Avg) | 3588.0 | 212.8 | 146.2 | 0.24166 (±0.00000) Rank: 12/51 |
0.35672 (±0.00000) Rank: 12/51 |
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) / ±0.12667 over 17 scenes)
Rank (per category): 16 (of 51)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| AMS | 4306.7 | 323.39 | 82.93 Rank: 15/51 |
1548.67 Rank: 19/51 |
3.309 Rank: 20/51 |
23.62568 Rank: 16/51 |
0.26053 (±0.00000) Rank: 16/51 |
0.34767 (±0.00000) Rank: 16/51 |
| BGK | 3087.8 | 293.83 | 94.35 Rank: 14/51 |
1979.45 Rank: 19/51 |
3.407 Rank: 18/51 |
13.14840 Rank: 15/51 |
0.02210 (±0.00000) Rank: 17/51 |
0.07467 (±0.00000) Rank: 18/51 |
| BCN | 4576.9 | 282.18 | 77.73 Rank: 20/51 |
1051.10 Rank: 19/51 |
3.310 Rank: 20/51 |
25.70677 Rank: 11/51 |
0.10849 (±0.00000) Rank: 15/51 |
0.17271 (±0.00000) Rank: 15/51 |
| BAR | 3994.0 | 298.75 | 70.13 Rank: 17/51 |
840.65 Rank: 20/51 |
3.011 Rank: 19/51 |
10.48137 Rank: 6/51 |
0.08984 (±0.00000) Rank: 15/51 |
0.14334 (±0.00000) Rank: 16/51 |
| CAM | 2981.6 | 210.23 | 74.19 Rank: 19/51 |
880.08 Rank: 19/51 |
2.941 Rank: 21/51 |
21.17913 Rank: 12/51 |
0.11837 (±0.00000) Rank: 15/51 |
0.16990 (±0.00000) Rank: 17/51 |
| CAN | 3345.9 | 283.45 | 66.10 Rank: 14/51 |
494.97 Rank: 22/51 |
3.167 Rank: 17/51 |
14.90292 Rank: 14/51 |
0.02270 (±0.00000) Rank: 19/51 |
0.04759 (±0.00000) Rank: 21/51 |
| CHI | 4274.9 | 268.08 | 85.60 Rank: 13/51 |
496.79 Rank: 16/51 |
3.998 Rank: 18/51 |
44.02110 Rank: 13/51 |
0.06821 (±0.00000) Rank: 18/51 |
0.11841 (±0.00000) Rank: 19/51 |
| HEL | 3348.7 | 391.04 | 91.28 Rank: 15/51 |
2194.88 Rank: 20/51 |
3.361 Rank: 12/51 |
12.70826 Rank: 15/51 |
0.26910 (±0.00000) Rank: 18/51 |
0.41956 (±0.00000) Rank: 15/51 |
| MAD | 3668.0 | 217.16 | 81.19 Rank: 15/51 |
951.90 Rank: 22/51 |
3.059 Rank: 21/51 |
5.63164 Rank: 19/51 |
0.03435 (±0.00000) Rank: 22/51 |
0.07885 (±0.00000) Rank: 22/51 |
| MTV | 3896.6 | 233.26 | 86.02 Rank: 13/51 |
1003.61 Rank: 21/51 |
3.295 Rank: 20/51 |
14.09408 Rank: 12/51 |
0.15476 (±0.00000) Rank: 15/51 |
0.24201 (±0.00000) Rank: 17/51 |
| NOR | 3534.6 | 234.21 | 68.14 Rank: 12/51 |
618.94 Rank: 19/51 |
2.984 Rank: 15/51 |
8.67398 Rank: 13/51 |
0.09247 (±0.00000) Rank: 17/51 |
0.12419 (±0.00000) Rank: 18/51 |
| SF | 3314.3 | 271.93 | 82.42 Rank: 13/51 |
1250.79 Rank: 17/51 |
3.162 Rank: 19/51 |
15.37122 Rank: 14/51 |
0.14946 (±0.00000) Rank: 15/51 |
0.24277 (±0.00000) Rank: 15/51 |
| SG | 2854.6 | 245.25 | 71.95 Rank: 11/51 |
1021.88 Rank: 14/51 |
3.123 Rank: 15/51 |
16.96041 Rank: 19/51 |
0.14593 (±0.00000) Rank: 8/51 |
0.20439 (±0.00000) Rank: 8/51 |
| SYD | 3081.6 | 229.51 | 71.94 Rank: 19/51 |
718.22 Rank: 20/51 |
2.973 Rank: 21/51 |
13.06313 Rank: 14/51 |
0.12914 (±0.00000) Rank: 18/51 |
0.19451 (±0.00000) Rank: 18/51 |
| TOK | 3047.8 | 286.01 | 86.49 Rank: 13/51 |
1671.79 Rank: 17/51 |
3.411 Rank: 11/51 |
12.42509 Rank: 17/51 |
0.48059 (±0.00000) Rank: 11/51 |
0.55092 (±0.00000) Rank: 11/51 |
| TOR | 3514.5 | 204.25 | 74.65 Rank: 17/51 |
617.33 Rank: 20/51 |
2.993 Rank: 20/51 |
11.90896 Rank: 9/51 |
0.07116 (±0.00000) Rank: 17/51 |
0.13020 (±0.00000) Rank: 16/51 |
| ZRH | 4168.4 | 307.54 | 82.24 Rank: 12/51 |
1859.69 Rank: 14/51 |
3.137 Rank: 9/51 |
17.94302 Rank: 17/51 |
0.10741 (±0.00000) Rank: 10/51 |
0.18199 (±0.00000) Rank: 10/51 |
| Avg | 3588.0 | 269.42 | 79.26 Rank: 14/51 |
1129.46 Rank: 19/51 |
3.214 Rank: 18/51 |
16.57913 Rank: 15/51 |
0.13674 (±0.00000) Rank: 15/51 |
0.20257 (±0.00000) Rank: 16/51 |
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 —
