Submission ID: 94de482d
KORNIA TUTORIAL AdaLAM CV-DoG-AffNet-HardNet8
Processed: 21-05-28. Download link: 94de482da7197801-dog-affnet-hardnet8-degensac-adalam.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: Dmytro Mishkin, Milan Pultar and kornia team (contact)
- Keypoint: cv2dog
- Descriptor: affnethardnet8 (128 float32: 512 bytes)
- Number of features: 8000
- Summary: OpeCV SIFT keypoints 8000 features, followed by the AffNet normalization and HardNet8 descriptor as implemented in kornia. Matched using AdaLAM with DEGENSAC
- Paper: https://arxiv.org/abs/2007.09699
- Website: https://github.com/kornia/kornia
- Processing date: 21-05-28
Phototourism dataset / Stereo track
mAA at 10 degrees: 0.55026 (±0.00000 over 1 run(s) / ±0.15095 over 9 scenes)
Rank (per category): 23 (of 51)
Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
BM | 7969.3 | — | 751.7 | 0.526 Rank: 38/51 |
0.892 Rank: 38/51 |
0.17226 (±0.00000) Rank: 35/51 |
0.26926 (±0.00000) Rank: 35/51 |
FCS | 7876.6 | — | 785.7 | 0.443 Rank: 39/51 |
0.929 Rank: 6/51 |
0.64292 (±0.00000) Rank: 19/51 |
0.75518 (±0.00000) Rank: 20/51 |
LMS | 7773.0 | — | 451.7 | 0.465 Rank: 28/51 |
0.685 Rank: 22/51 |
0.54660 (±0.00000) Rank: 24/51 |
0.65421 (±0.00000) Rank: 26/51 |
LB | 7958.1 | — | 432.8 | 0.472 Rank: 37/51 |
0.765 Rank: 14/51 |
0.47821 (±0.00000) Rank: 35/51 |
0.56495 (±0.00000) Rank: 26/51 |
MC | 7828.2 | — | 690.5 | 0.525 Rank: 43/51 |
0.948 Rank: 9/51 |
0.44710 (±0.00000) Rank: 35/51 |
0.59267 (±0.00000) Rank: 5/51 |
MR | 7795.9 | — | 657.3 | 0.457 Rank: 43/51 |
0.966 Rank: 5/51 |
0.35553 (±0.00000) Rank: 3/51 |
0.47217 (±0.00000) Rank: 4/51 |
PSM | 7811.8 | — | 385.0 | 0.385 Rank: 46/51 |
0.769 Rank: 7/51 |
0.20201 (±0.00000) Rank: 13/51 |
0.34009 (±0.00000) Rank: 14/51 |
SF | 7900.7 | — | 792.5 | 0.479 Rank: 36/51 |
0.879 Rank: 4/51 |
0.54160 (±0.00000) Rank: 19/51 |
0.66530 (±0.00000) Rank: 19/51 |
SPC | 7871.1 | — | 681.8 | 0.487 Rank: 40/51 |
0.876 Rank: 5/51 |
0.48182 (±0.00000) Rank: 20/51 |
0.63853 (±0.00000) Rank: 17/51 |
Avg | 7865.0 | — | 625.4 | 0.471 Rank: 43/51 |
0.857 Rank: 6/51 |
0.42978 (±0.00000) Rank: 19/51 |
0.55026 (±0.00000) Rank: 23/51 |
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) / ±0.11033 over 9 scenes)
Rank (per category): 17 (of 51)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
BM | 7969.3 | 749.81 | 99.87 Rank: 14/51 |
4218.52 Rank: 20/51 |
5.195 Rank: 18/51 |
0.50664 Rank: 34/51 |
0.42901 (±0.00000) Rank: 37/51 |
0.58657 (±0.00000) Rank: 35/51 |
FCS | 7876.6 | 819.68 | 99.01 Rank: 7/51 |
5910.67 Rank: 19/51 |
4.781 Rank: 22/51 |
0.20715 Rank: 3/51 |
0.77990 (±0.00000) Rank: 3/51 |
0.83558 (±0.00000) Rank: 3/51 |
LMS | 7773.0 | 489.21 | 99.07 Rank: 29/51 |
4302.36 Rank: 20/51 |
4.588 Rank: 27/51 |
0.27803 Rank: 8/51 |
0.82943 (±0.00000) Rank: 17/51 |
0.87872 (±0.00000) Rank: 21/51 |
LB | 7958.1 | 527.83 | 98.36 Rank: 1/51 |
3859.54 Rank: 23/51 |
4.765 Rank: 26/51 |
0.45341 Rank: 2/51 |
0.75096 (±0.00000) Rank: 6/51 |
0.82943 (±0.00000) Rank: 4/51 |
MC | 7828.2 | 677.80 | 100.00 Rank: 1/51 |
4856.07 Rank: 21/51 |
4.976 Rank: 28/51 |
0.33031 Rank: 14/51 |
0.57511 (±0.00000) Rank: 9/51 |
0.71703 (±0.00000) Rank: 7/51 |
MR | 7795.9 | 659.34 | 92.67 Rank: 39/51 |
5263.16 Rank: 25/51 |
4.220 Rank: 36/51 |
0.47718 Rank: 15/51 |
0.46503 (±0.00000) Rank: 17/51 |
0.57450 (±0.00000) Rank: 19/51 |
PSM | 7811.8 | 378.28 | 98.52 Rank: 18/51 |
4757.25 Rank: 26/51 |
3.716 Rank: 26/51 |
0.40480 Rank: 15/51 |
0.58958 (±0.00000) Rank: 14/51 |
0.67469 (±0.00000) Rank: 14/51 |
SF | 7900.7 | 780.46 | 99.43 Rank: 18/51 |
6353.54 Rank: 24/51 |
4.739 Rank: 26/51 |
0.29833 Rank: 19/51 |
0.75339 (±0.00000) Rank: 20/51 |
0.83863 (±0.00000) Rank: 20/51 |
SPC | 7871.1 | 709.38 | 100.00 Rank: 1/51 |
5427.73 Rank: 17/51 |
4.854 Rank: 22/51 |
0.37401 Rank: 1/51 |
0.74242 (±0.00000) Rank: 11/51 |
0.83532 (±0.00000) Rank: 13/51 |
Avg | 7865.0 | 643.53 | 98.55 Rank: 22/51 |
4994.32 Rank: 21/51 |
4.648 Rank: 26/51 |
0.36998 Rank: 8/51 |
0.65720 (±0.00000) Rank: 15/51 |
0.75227 (±0.00000) Rank: 17/51 |
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: 0.59985 (±0.00000 over 1 run(s) / ±0.05754 over 3 scenes)
Rank (per category): 24 (of 51)
Scene | Features | Matches (raw) |
Matches (final) |
Rep. @ 3 px. | MS @ 3 px. | mAA(5o) | mAA(10o) |
Lizard | 8000.0 | — | 559.0 | 0.121 Rank: 20/51 |
0.012 Rank: 8/51 |
0.39778 (±0.00000) Rank: 24/51 |
0.52843 (±0.00000) Rank: 24/51 |
Pond | 8000.0 | — | 649.4 | 0.152 Rank: 18/51 |
0.055 Rank: 15/51 |
0.53507 (±0.00000) Rank: 19/51 |
0.66934 (±0.00000) Rank: 18/51 |
Tree | 8000.0 | — | 370.4 | 0.106 Rank: 27/51 |
0.019 Rank: 27/51 |
0.50155 (±0.00000) Rank: 29/51 |
0.60177 (±0.00000) Rank: 29/51 |
Avg | 8000.0 | — | 526.3 | 0.126 Rank: 18/51 |
0.029 Rank: 7/51 |
0.47813 (±0.00000) Rank: 24/51 |
0.59985 (±0.00000) Rank: 24/51 |
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) / ±0.14933 over 3 scenes)
Rank (per category): 18 (of 51)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
Lizard | 8000.0 | 616.68 | 84.17 Rank: 19/51 |
1737.82 Rank: 27/51 |
3.227 Rank: 24/51 |
13.73332 Rank: 8/51 |
0.55033 (±0.00000) Rank: 20/51 |
0.63870 (±0.00000) Rank: 18/51 |
Pond | 8000.0 | 864.50 | 68.70 Rank: 24/51 |
2109.59 Rank: 23/51 |
2.898 Rank: 24/51 |
0.42417 Rank: 13/51 |
0.24707 (±0.00000) Rank: 30/51 |
0.27607 (±0.00000) Rank: 33/51 |
Tree | 8000.0 | 501.59 | 73.21 Rank: 8/51 |
1183.11 Rank: 20/51 |
2.944 Rank: 21/51 |
5.62085 Rank: 7/51 |
0.46116 (±0.00000) Rank: 13/51 |
0.49886 (±0.00000) Rank: 11/51 |
Avg | 8000.0 | 660.92 | 75.36 Rank: 16/51 |
1676.84 Rank: 25/51 |
3.023 Rank: 26/51 |
6.59278 Rank: 9/51 |
0.41952 (±0.00000) Rank: 21/51 |
0.47121 (±0.00000) Rank: 18/51 |
Google Urban dataset / Stereo track
mAA at 10 degrees: 0.29371 (±0.00000 over 1 run(s) / ±0.11588 over 17 scenes)
Rank (per category): 20 (of 51)
Scene | Features | Matches (raw) |
Matches (final) |
mAA(5o) | mAA(10o) |
Amsterdam (AMS) | 6126.2 | — | 259.2 | 0.22411 (±0.00000) Rank: 16/51 |
0.35491 (±0.00000) Rank: 15/51 |
Bangkok (BGK) | 6403.9 | — | 282.9 | 0.05707 (±0.00000) Rank: 20/51 |
0.16030 (±0.00000) Rank: 19/51 |
Barcelona (BCN) | 5657.4 | — | 154.5 | 0.11355 (±0.00000) Rank: 16/51 |
0.19341 (±0.00000) Rank: 16/51 |
Buenos Aires (BAR) | 6586.7 | — | 233.5 | 0.17900 (±0.00000) Rank: 16/51 |
0.30137 (±0.00000) Rank: 18/51 |
Cambridge (CAM) | 6200.9 | — | 170.2 | 0.20479 (±0.00000) Rank: 16/51 |
0.32603 (±0.00000) Rank: 21/51 |
Cannes (CAN) | 5862.6 | — | 197.5 | 0.30323 (±0.00000) Rank: 24/51 |
0.40461 (±0.00000) Rank: 22/51 |
Chicago (CHI) | 5709.0 | — | 64.5 | 0.05383 (±0.00000) Rank: 22/51 |
0.10260 (±0.00000) Rank: 23/51 |
Helsinki (HEL) | 6653.0 | — | 399.5 | 0.33918 (±0.00000) Rank: 20/51 |
0.52680 (±0.00000) Rank: 20/51 |
Madrid (MAD) | 7031.3 | — | 191.4 | 0.06079 (±0.00000) Rank: 17/51 |
0.15380 (±0.00000) Rank: 14/51 |
Mountain View (MTV) | 6280.8 | — | 139.6 | 0.14297 (±0.00000) Rank: 18/51 |
0.23223 (±0.00000) Rank: 19/51 |
New Orleans (NOR) | 6235.4 | — | 147.0 | 0.18521 (±0.00000) Rank: 21/51 |
0.24260 (±0.00000) Rank: 22/51 |
San Francisco (SF) | 6356.8 | — | 195.1 | 0.11864 (±0.00000) Rank: 19/51 |
0.21339 (±0.00000) Rank: 21/51 |
Singapore (SG) | 7303.6 | — | 196.2 | 0.24241 (±0.00000) Rank: 15/51 |
0.34114 (±0.00000) Rank: 15/51 |
Sydney (SYD) | 6676.7 | — | 194.7 | 0.23493 (±0.00000) Rank: 18/51 |
0.35415 (±0.00000) Rank: 19/51 |
Tokyo (TOK) | 6592.6 | — | 235.2 | 0.40196 (±0.00000) Rank: 19/51 |
0.52010 (±0.00000) Rank: 21/51 |
Toronto (TOR) | 7153.6 | — | 117.4 | 0.13200 (±0.00000) Rank: 20/51 |
0.24680 (±0.00000) Rank: 20/51 |
Zurich (ZRH) | 6010.7 | — | 241.7 | 0.22680 (±0.00000) Rank: 19/51 |
0.31890 (±0.00000) Rank: 16/51 |
Average (Avg) | 6402.4 | — | 201.2 | 0.18944 (±0.00000) Rank: 20/51 |
0.29371 (±0.00000) Rank: 20/51 |
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) / ±0.12853 over 17 scenes)
Rank (per category): 20 (of 51)
Scene | Features | Matches (input) |
RegistrationRatio (%) | Number of Landmarks |
Track Length | ATE | mAA(50) | mAA(100) |
AMS | 6126.2 | 152.54 | 77.73 Rank: 24/51 |
1384.85 Rank: 23/51 |
3.229 Rank: 23/51 |
20.86311 Rank: 8/51 |
0.24265 (±0.00000) Rank: 19/51 |
0.31072 (±0.00000) Rank: 22/51 |
BGK | 6403.9 | 182.70 | 90.80 Rank: 24/51 |
1902.45 Rank: 20/51 |
3.329 Rank: 21/51 |
14.33484 Rank: 20/51 |
0.02541 (±0.00000) Rank: 11/51 |
0.07657 (±0.00000) Rank: 16/51 |
BCN | 5657.4 | 89.14 | 75.33 Rank: 23/51 |
600.46 Rank: 25/51 |
3.344 Rank: 13/51 |
24.10008 Rank: 7/51 |
0.07513 (±0.00000) Rank: 23/51 |
0.11871 (±0.00000) Rank: 23/51 |
BAR | 6586.7 | 119.14 | 66.64 Rank: 21/51 |
743.41 Rank: 23/51 |
2.986 Rank: 20/51 |
12.26654 Rank: 11/51 |
0.06751 (±0.00000) Rank: 18/51 |
0.12330 (±0.00000) Rank: 18/51 |
CAM | 6200.9 | 95.65 | 69.82 Rank: 23/51 |
805.31 Rank: 23/51 |
2.882 Rank: 23/51 |
14.53807 Rank: 4/51 |
0.08833 (±0.00000) Rank: 24/51 |
0.13904 (±0.00000) Rank: 23/51 |
CAN | 5862.6 | 102.80 | 52.62 Rank: 24/51 |
357.86 Rank: 25/51 |
2.935 Rank: 21/51 |
7.16591 Rank: 3/51 |
0.00787 (±0.00000) Rank: 26/51 |
0.02613 (±0.00000) Rank: 25/51 |
CHI | 5709.0 | 57.11 | 57.40 Rank: 24/51 |
137.91 Rank: 26/51 |
3.258 Rank: 23/51 |
29.74583 Rank: 3/51 |
0.01242 (±0.00000) Rank: 23/51 |
0.02881 (±0.00000) Rank: 24/51 |
HEL | 6653.0 | 281.19 | 93.06 Rank: 11/51 |
2659.25 Rank: 17/51 |
3.226 Rank: 23/51 |
14.51773 Rank: 27/51 |
0.30273 (±0.00000) Rank: 6/51 |
0.45872 (±0.00000) Rank: 1/51 |
MAD | 7031.3 | 115.52 | 80.91 Rank: 17/51 |
1266.69 Rank: 19/51 |
3.078 Rank: 20/51 |
5.46486 Rank: 16/51 |
0.04307 (±0.00000) Rank: 15/51 |
0.09629 (±0.00000) Rank: 16/51 |
MTV | 6280.8 | 96.03 | 80.25 Rank: 23/51 |
1025.84 Rank: 18/51 |
3.214 Rank: 21/51 |
13.74804 Rank: 10/51 |
0.13291 (±0.00000) Rank: 18/51 |
0.21843 (±0.00000) Rank: 20/51 |
NOR | 6235.4 | 93.64 | 54.58 Rank: 23/51 |
367.54 Rank: 25/51 |
2.766 Rank: 21/51 |
5.85816 Rank: 5/51 |
0.05575 (±0.00000) Rank: 22/51 |
0.08505 (±0.00000) Rank: 23/51 |
SF | 6356.8 | 109.69 | 75.24 Rank: 22/51 |
804.38 Rank: 25/51 |
3.087 Rank: 20/51 |
12.33402 Rank: 7/51 |
0.10769 (±0.00000) Rank: 21/51 |
0.17191 (±0.00000) Rank: 21/51 |
SG | 7303.6 | 116.21 | 60.78 Rank: 22/51 |
682.42 Rank: 22/51 |
2.723 Rank: 27/51 |
10.72361 Rank: 4/51 |
0.07616 (±0.00000) Rank: 18/51 |
0.10870 (±0.00000) Rank: 21/51 |
SYD | 6676.7 | 107.19 | 69.40 Rank: 22/51 |
726.03 Rank: 19/51 |
3.060 Rank: 17/51 |
12.86425 Rank: 12/51 |
0.14097 (±0.00000) Rank: 17/51 |
0.19502 (±0.00000) Rank: 17/51 |
TOK | 6592.6 | 141.32 | 81.90 Rank: 20/51 |
1522.98 Rank: 19/51 |
3.290 Rank: 17/51 |
9.81665 Rank: 8/51 |
0.40209 (±0.00000) Rank: 20/51 |
0.46800 (±0.00000) Rank: 19/51 |
TOR | 7153.6 | 65.12 | 69.73 Rank: 23/51 |
575.70 Rank: 21/51 |
3.000 Rank: 19/51 |
11.78440 Rank: 8/51 |
0.07156 (±0.00000) Rank: 16/51 |
0.12749 (±0.00000) Rank: 17/51 |
ZRH | 6010.7 | 135.22 | 69.68 Rank: 21/51 |
1040.43 Rank: 24/51 |
2.937 Rank: 22/51 |
13.60970 Rank: 9/51 |
0.02613 (±0.00000) Rank: 23/51 |
0.05627 (±0.00000) Rank: 23/51 |
Avg | 6402.4 | 121.19 | 72.11 Rank: 22/51 |
976.68 Rank: 22/51 |
3.079 Rank: 21/51 |
13.74916 Rank: 5/51 |
0.11049 (±0.00000) Rank: 20/51 |
0.16524 (±0.00000) Rank: 20/51 |
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 —









