Submission ID: 1bca437f
HarrisZ improved AffNet-HardNet8
Processed: 21-06-12. Download link: 1bca437f151e5bd1-harrisz-2k.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: hzbnn22l05m2k2l
- Descriptor: hardnet (128 float32: 512 bytes)
- Number of features: 2048
- 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: 21-06-12
Phototourism dataset / Stereo track
mAA at 10 degrees: 0.47539 (±0.00000 over 1 run(s) / ±0.14287 over 9 scenes)
Rank (per category): 51 (of 87)
| Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| BM | 2048.0 | 188.9 | 141.2 | 0.470 Rank: 32/87 |
0.952 Rank: 5/87 |
0.21850 (±0.00000) Rank: 55/87 |
0.34346 (±0.00000) Rank: 55/87 |
| FCS | 2048.0 | 189.3 | 147.4 | 0.348 Rank: 79/87 |
0.928 Rank: 1/87 |
0.56712 (±0.00000) Rank: 46/87 |
0.69101 (±0.00000) Rank: 46/87 |
| LMS | 2025.2 | 169.2 | 112.1 | 0.329 Rank: 83/87 |
0.680 Rank: 45/87 |
0.50610 (±0.00000) Rank: 52/87 |
0.64112 (±0.00000) Rank: 51/87 |
| LB | 2046.5 | 153.9 | 103.5 | 0.336 Rank: 76/87 |
0.775 Rank: 3/87 |
0.39956 (±0.00000) Rank: 55/87 |
0.51288 (±0.00000) Rank: 52/87 |
| MC | 2047.5 | 180.9 | 141.3 | 0.401 Rank: 56/87 |
0.952 Rank: 3/87 |
0.30437 (±0.00000) Rank: 55/87 |
0.45126 (±0.00000) Rank: 49/87 |
| MR | 2048.0 | 208.3 | 161.4 | 0.355 Rank: 81/87 |
0.953 Rank: 2/87 |
0.23999 (±0.00000) Rank: 43/87 |
0.34365 (±0.00000) Rank: 44/87 |
| PSM | 2048.0 | 95.5 | 70.4 | 0.289 Rank: 76/87 |
0.697 Rank: 5/87 |
0.12142 (±0.00000) Rank: 55/87 |
0.22240 (±0.00000) Rank: 56/87 |
| SF | 2048.0 | 212.8 | 166.3 | 0.320 Rank: 82/87 |
0.871 Rank: 1/87 |
0.42883 (±0.00000) Rank: 48/87 |
0.56432 (±0.00000) Rank: 48/87 |
| SPC | 2043.3 | 149.3 | 101.5 | 0.344 Rank: 72/87 |
0.864 Rank: 3/87 |
0.36237 (±0.00000) Rank: 52/87 |
0.50838 (±0.00000) Rank: 55/87 |
| Avg | 2044.7 | 172.0 | 127.3 | 0.355 Rank: 75/87 |
0.853 Rank: 1/87 |
0.34981 (±0.00000) Rank: 51/87 |
0.47539 (±0.00000) Rank: 51/87 |
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) / ±0.14652 over 9 scenes)
Rank (per category): 62 (of 87)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| BM | 2048.0 | 288.27 | 99.93 Rank: 24/87 |
1218.66 Rank: 59/87 |
4.440 Rank: 76/87 |
0.54526 Rank: 61/87 |
0.41879 (±0.00000) Rank: 62/87 |
0.55695 (±0.00000) Rank: 63/87 |
| FCS | 2048.0 | 299.70 | 98.33 Rank: 16/87 |
1769.17 Rank: 63/87 |
4.278 Rank: 65/87 |
0.23660 Rank: 20/87 |
0.74250 (±0.00000) Rank: 29/87 |
0.80899 (±0.00000) Rank: 16/87 |
| LMS | 2025.2 | 289.05 | 98.53 Rank: 68/87 |
1321.17 Rank: 66/87 |
4.494 Rank: 77/87 |
0.28200 Rank: 25/87 |
0.81559 (±0.00000) Rank: 50/87 |
0.87493 (±0.00000) Rank: 53/87 |
| LB | 2046.5 | 283.00 | 97.65 Rank: 59/87 |
1175.39 Rank: 62/87 |
4.358 Rank: 72/87 |
0.52207 Rank: 72/87 |
0.61946 (±0.00000) Rank: 63/87 |
0.71941 (±0.00000) Rank: 64/87 |
| MC | 2047.5 | 275.50 | 99.40 Rank: 63/87 |
1367.97 Rank: 60/87 |
4.385 Rank: 71/87 |
0.45067 Rank: 67/87 |
0.47493 (±0.00000) Rank: 62/87 |
0.61737 (±0.00000) Rank: 64/87 |
| MR | 2048.0 | 293.92 | 91.67 Rank: 67/87 |
1488.07 Rank: 57/87 |
4.101 Rank: 76/87 |
0.61042 Rank: 71/87 |
0.37062 (±0.00000) Rank: 57/87 |
0.48154 (±0.00000) Rank: 61/87 |
| PSM | 2048.0 | 188.30 | 92.06 Rank: 62/87 |
1716.43 Rank: 66/87 |
3.093 Rank: 70/87 |
0.65592 Rank: 66/87 |
0.36501 (±0.00000) Rank: 64/87 |
0.45524 (±0.00000) Rank: 66/87 |
| SF | 2048.0 | 302.24 | 98.15 Rank: 51/87 |
1837.46 Rank: 62/87 |
4.237 Rank: 77/87 |
0.30088 Rank: 45/87 |
0.73075 (±0.00000) Rank: 48/87 |
0.81627 (±0.00000) Rank: 49/87 |
| SPC | 2043.3 | 257.91 | 99.06 Rank: 61/87 |
1394.85 Rank: 62/87 |
4.144 Rank: 75/87 |
0.46972 Rank: 46/87 |
0.67343 (±0.00000) Rank: 62/87 |
0.77749 (±0.00000) Rank: 63/87 |
| Avg | 2044.7 | 275.32 | 97.20 Rank: 61/87 |
1476.58 Rank: 60/87 |
4.170 Rank: 76/87 |
0.45262 Rank: 61/87 |
0.57901 (±0.00000) Rank: 62/87 |
0.67869 (±0.00000) Rank: 62/87 |
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: 0.63586 (±0.00000 over 1 run(s) / ±0.07866 over 3 scenes)
Rank (per category): 42 (of 87)
| Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
| Lizard | 2048.0 | 180.6 | 168.7 | 0.024 Rank: 68/87 |
0.010 Rank: 33/87 |
0.47795 (±0.00000) Rank: 41/87 |
0.60555 (±0.00000) Rank: 41/87 |
| Pond | 2048.0 | 164.8 | 150.1 | 0.047 Rank: 58/87 |
0.051 Rank: 20/87 |
0.43487 (±0.00000) Rank: 49/87 |
0.55832 (±0.00000) Rank: 50/87 |
| Tree | 2048.0 | 120.7 | 107.4 | 0.029 Rank: 53/87 |
0.025 Rank: 20/87 |
0.65033 (±0.00000) Rank: 38/87 |
0.74371 (±0.00000) Rank: 42/87 |
| Avg | 2048.0 | 155.4 | 142.1 | 0.034 Rank: 68/87 |
0.029 Rank: 9/87 |
0.52105 (±0.00000) Rank: 42/87 |
0.63586 (±0.00000) Rank: 42/87 |
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) / ±0.16484 over 3 scenes)
Rank (per category): 45 (of 87)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| Lizard | 2048.0 | 287.53 | 85.40 Rank: 18/87 |
694.67 Rank: 54/87 |
3.376 Rank: 39/87 |
14.38787 Rank: 54/87 |
0.56873 (±0.00000) Rank: 41/87 |
0.65674 (±0.00000) Rank: 41/87 |
| Pond | 2048.0 | 342.69 | 70.23 Rank: 12/87 |
636.45 Rank: 64/87 |
2.990 Rank: 22/87 |
0.44704 Rank: 67/87 |
0.23422 (±0.00000) Rank: 63/87 |
0.25504 (±0.00000) Rank: 65/87 |
| Tree | 2048.0 | 213.61 | 71.81 Rank: 24/87 |
344.98 Rank: 64/87 |
3.015 Rank: 22/87 |
5.68825 Rank: 7/87 |
0.45158 (±0.00000) Rank: 18/87 |
0.49126 (±0.00000) Rank: 15/87 |
| Avg | 2048.0 | 281.28 | 75.82 Rank: 9/87 |
558.70 Rank: 63/87 |
3.127 Rank: 20/87 |
6.84105 Rank: 25/87 |
0.41818 (±0.00000) Rank: 43/87 |
0.46768 (±0.00000) Rank: 45/87 |
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.
— Lizard —
— Pond —
— Tree —
Google Urban dataset / Stereo track
mAA at 10 degrees: 0.32526 (±0.00000 over 1 run(s) / ±0.13716 over 17 scenes)
Rank (per category): 39 (of 87)
| Scene | Features | Matches (raw) |
Matches (final) |
mAA(5o) | mAA(10o) |
| Amsterdam (AMS) | 2048.0 | 117.3 | 87.9 | 0.16429 (±0.00000) Rank: 40/87 |
0.27634 (±0.00000) Rank: 42/87 |
| Bangkok (BGK) | 2026.9 | 149.4 | 104.7 | 0.05161 (±0.00000) Rank: 37/87 |
0.14169 (±0.00000) Rank: 42/87 |
| Barcelona (BCN) | 2048.0 | 100.9 | 75.6 | 0.14139 (±0.00000) Rank: 7/87 |
0.20220 (±0.00000) Rank: 29/87 |
| Buenos Aires (BAR) | 2048.0 | 135.7 | 98.4 | 0.16621 (±0.00000) Rank: 40/87 |
0.28174 (±0.00000) Rank: 43/87 |
| Cambridge (CAM) | 2044.9 | 107.3 | 76.1 | 0.27603 (±0.00000) Rank: 40/87 |
0.40377 (±0.00000) Rank: 35/87 |
| Cannes (CAN) | 2048.0 | 154.4 | 108.8 | 0.40184 (±0.00000) Rank: 20/87 |
0.50092 (±0.00000) Rank: 31/87 |
| Chicago (CHI) | 2042.0 | 67.6 | 46.6 | 0.18934 (±0.00000) Rank: 31/87 |
0.29433 (±0.00000) Rank: 31/87 |
| Helsinki (HEL) | 2043.3 | 206.1 | 144.0 | 0.36907 (±0.00000) Rank: 28/87 |
0.57113 (±0.00000) Rank: 27/87 |
| Madrid (MAD) | 2046.3 | 75.8 | 47.2 | 0.02188 (±0.00000) Rank: 48/87 |
0.08267 (±0.00000) Rank: 48/87 |
| Mountain View (MTV) | 2047.9 | 74.2 | 49.4 | 0.15391 (±0.00000) Rank: 42/87 |
0.25449 (±0.00000) Rank: 41/87 |
| New Orleans (NOR) | 2038.9 | 101.9 | 70.0 | 0.21953 (±0.00000) Rank: 38/87 |
0.29793 (±0.00000) Rank: 38/87 |
| San Francisco (SF) | 2041.1 | 136.4 | 103.5 | 0.16010 (±0.00000) Rank: 38/87 |
0.27638 (±0.00000) Rank: 40/87 |
| Singapore (SG) | 2033.1 | 150.8 | 102.3 | 0.28608 (±0.00000) Rank: 36/87 |
0.40443 (±0.00000) Rank: 36/87 |
| Sydney (SYD) | 2019.0 | 124.4 | 82.7 | 0.27074 (±0.00000) Rank: 37/87 |
0.40000 (±0.00000) Rank: 38/87 |
| Tokyo (TOK) | 2039.4 | 176.7 | 112.8 | 0.46912 (±0.00000) Rank: 41/87 |
0.60588 (±0.00000) Rank: 41/87 |
| Toronto (TOR) | 2041.5 | 69.5 | 43.4 | 0.12000 (±0.00000) Rank: 46/87 |
0.22280 (±0.00000) Rank: 46/87 |
| Zurich (ZRH) | 2043.6 | 123.7 | 92.2 | 0.24192 (±0.00000) Rank: 29/87 |
0.31271 (±0.00000) Rank: 37/87 |
| Average (Avg) | 2041.2 | 121.9 | 85.0 | 0.21783 (±0.00000) Rank: 37/87 |
0.32526 (±0.00000) Rank: 39/87 |
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) / ±0.12325 over 17 scenes)
Rank (per category): 46 (of 87)
| Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
| AMS | 2048.0 | 150.26 | 67.53 Rank: 47/87 |
584.54 Rank: 47/87 |
3.052 Rank: 50/87 |
15.98680 Rank: 3/87 |
0.13544 (±0.00000) Rank: 49/87 |
0.19279 (±0.00000) Rank: 47/87 |
| BGK | 2026.9 | 190.63 | 91.30 Rank: 46/87 |
1260.58 Rank: 42/87 |
3.347 Rank: 51/87 |
12.25063 Rank: 16/87 |
0.02087 (±0.00000) Rank: 45/87 |
0.07137 (±0.00000) Rank: 42/87 |
| BCN | 2048.0 | 132.37 | 69.12 Rank: 48/87 |
448.26 Rank: 48/87 |
3.091 Rank: 52/87 |
21.34386 Rank: 3/87 |
0.07314 (±0.00000) Rank: 46/87 |
0.11679 (±0.00000) Rank: 47/87 |
| BAR | 2048.0 | 157.48 | 59.16 Rank: 48/87 |
422.81 Rank: 46/87 |
2.791 Rank: 50/87 |
9.32928 Rank: 5/87 |
0.06252 (±0.00000) Rank: 45/87 |
0.10896 (±0.00000) Rank: 46/87 |
| CAM | 2044.9 | 148.42 | 65.55 Rank: 49/87 |
621.71 Rank: 44/87 |
2.827 Rank: 51/87 |
14.31496 Rank: 4/87 |
0.09344 (±0.00000) Rank: 48/87 |
0.13595 (±0.00000) Rank: 49/87 |
| CAN | 2048.0 | 170.94 | 53.24 Rank: 51/87 |
255.61 Rank: 50/87 |
2.871 Rank: 51/87 |
12.53564 Rank: 9/87 |
0.03346 (±0.00000) Rank: 39/87 |
0.05184 (±0.00000) Rank: 44/87 |
| CHI | 2042.0 | 137.59 | 76.72 Rank: 42/87 |
282.78 Rank: 41/87 |
3.835 Rank: 45/87 |
45.31726 Rank: 15/87 |
0.05947 (±0.00000) Rank: 37/87 |
0.11773 (±0.00000) Rank: 37/87 |
| HEL | 2043.3 | 241.76 | 89.60 Rank: 45/87 |
1476.84 Rank: 42/87 |
3.332 Rank: 51/87 |
11.21106 Rank: 8/87 |
0.27196 (±0.00000) Rank: 41/87 |
0.40920 (±0.00000) Rank: 41/87 |
| MAD | 2046.3 | 130.65 | 74.86 Rank: 46/87 |
525.59 Rank: 46/87 |
3.023 Rank: 51/87 |
4.89788 Rank: 15/87 |
0.02690 (±0.00000) Rank: 46/87 |
0.06255 (±0.00000) Rank: 46/87 |
| MTV | 2047.9 | 133.76 | 74.58 Rank: 48/87 |
552.55 Rank: 49/87 |
3.075 Rank: 51/87 |
11.65850 Rank: 6/87 |
0.12845 (±0.00000) Rank: 46/87 |
0.19770 (±0.00000) Rank: 46/87 |
| NOR | 2038.9 | 147.73 | 55.55 Rank: 49/87 |
307.93 Rank: 47/87 |
2.835 Rank: 49/87 |
6.60817 Rank: 6/87 |
0.05987 (±0.00000) Rank: 47/87 |
0.08545 (±0.00000) Rank: 48/87 |
| SF | 2041.1 | 168.02 | 73.85 Rank: 48/87 |
731.11 Rank: 45/87 |
3.048 Rank: 51/87 |
13.91122 Rank: 7/87 |
0.11021 (±0.00000) Rank: 46/87 |
0.17476 (±0.00000) Rank: 47/87 |
| SG | 2033.1 | 181.40 | 64.84 Rank: 44/87 |
766.64 Rank: 33/87 |
3.000 Rank: 50/87 |
14.55486 Rank: 10/87 |
0.11436 (±0.00000) Rank: 41/87 |
0.15484 (±0.00000) Rank: 41/87 |
| SYD | 2019.0 | 161.44 | 69.89 Rank: 45/87 |
547.54 Rank: 44/87 |
2.937 Rank: 49/87 |
9.03655 Rank: 5/87 |
0.11036 (±0.00000) Rank: 46/87 |
0.16143 (±0.00000) Rank: 46/87 |
| TOK | 2039.4 | 198.16 | 84.13 Rank: 44/87 |
1234.99 Rank: 43/87 |
3.272 Rank: 51/87 |
11.41349 Rank: 9/87 |
0.46493 (±0.00000) Rank: 44/87 |
0.53114 (±0.00000) Rank: 44/87 |
| TOR | 2041.5 | 126.83 | 62.06 Rank: 48/87 |
379.29 Rank: 45/87 |
2.838 Rank: 50/87 |
9.97217 Rank: 7/87 |
0.03464 (±0.00000) Rank: 47/87 |
0.07071 (±0.00000) Rank: 47/87 |
| ZRH | 2043.6 | 151.37 | 69.87 Rank: 47/87 |
831.06 Rank: 45/87 |
2.898 Rank: 51/87 |
15.22720 Rank: 13/87 |
0.04122 (±0.00000) Rank: 42/87 |
0.08107 (±0.00000) Rank: 42/87 |
| Avg | 2041.2 | 160.52 | 70.70 Rank: 47/87 |
660.58 Rank: 46/87 |
3.063 Rank: 51/87 |
14.09233 Rank: 7/87 |
0.10831 (±0.00000) Rank: 45/87 |
0.16025 (±0.00000) Rank: 46/87 |
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 —
