AMIR15: Online Deep Tracking Metric learning
TUD-Crossing
Benchmark:
Short name:
AMIR15
Detector:
Public
Description:
n/a
Reference:
A. Sadeghian, A. Alahi, S. Savarese. Tracking The Untrackable: Learning To Track Multiple Cues with Long-Term Dependencies. In ICCV, 2017.
Last submitted:
March 14, 2017 (3 years ago)
Published:
November 21, 2016 at 23:48:33 CET
Submissions:
3
Project page / code:
Open source:
No
Hardware:
TITAN X 3GHZ 1 Core
Runtime:
1.9 Hz
Benchmark performance:
| Sequence | MOTA | IDF1 | MOTP | MT | ML | FP | FN | Recall | Precision | FAF | ID Sw. | Frag |
| 2D MOT 2015 | 37.6 | 46.0 | 71.7 | 114 (15.8) | 193 (26.8) | 7,933 | 29,397 | 52.2 | 80.2 | 1.4 | 1,026 (19.7) | 2,024 (38.8) |
Detailed performance:
| Sequence | MOTA | IDF1 | MOTP | MT | ML | FP | FN | Recall | Precision | FAF | ID Sw. | Frag |
| ADL-Rundle-1 | 24.9 | 44.3 | 71.9 | 4 | 1 | 3,041 | 3,838 | 58.8 | 64.3 | 6.1 | 107 | 254 |
| ADL-Rundle-3 | 40.6 | 45.0 | 72.9 | 4 | 11 | 552 | 5,387 | 47.0 | 89.6 | 0.9 | 101 | 130 |
| AVG-TownCentre | 36.2 | 52.5 | 69.5 | 59 | 40 | 1,448 | 2,882 | 59.7 | 74.7 | 3.2 | 234 | 389 |
| ETH-Crossing | 32.9 | 46.8 | 76.3 | 2 | 10 | 33 | 619 | 38.3 | 92.1 | 0.2 | 21 | 47 |
| ETH-Jelmoli | 45.1 | 58.2 | 73.7 | 8 | 17 | 261 | 1,114 | 56.1 | 84.5 | 0.6 | 18 | 61 |
| ETH-Linthescher | 34.5 | 44.6 | 75.2 | 15 | 92 | 330 | 5,370 | 39.9 | 91.5 | 0.3 | 145 | 332 |
| KITTI-16 | 50.4 | 63.5 | 73.3 | 2 | 2 | 111 | 714 | 58.0 | 89.9 | 0.5 | 18 | 35 |
| KITTI-19 | 36.7 | 51.7 | 66.0 | 5 | 12 | 726 | 2,602 | 51.3 | 79.1 | 0.7 | 55 | 220 |
| PETS09-S2L2 | 47.0 | 36.3 | 70.5 | 5 | 4 | 616 | 4,236 | 56.1 | 89.8 | 1.4 | 254 | 397 |
| TUD-Crossing | 73.7 | 57.4 | 74.8 | 7 | 0 | 18 | 256 | 76.8 | 97.9 | 0.1 | 16 | 36 |
| Venice-1 | 29.1 | 40.4 | 72.5 | 3 | 4 | 797 | 2,379 | 47.9 | 73.3 | 1.8 | 57 | 123 |
Raw data:
