Appearance Model with R-CNN
Short name:
AP_HWDPL_p
Benchmark:
Description:
Online tracking with appearance model based on R-CNN from Huawei (HWDPL).
Hardware:
GTX 1080, 2.3 GHz, 1 Core
Detector:
Public
Processing:
Online
Last submitted:
December 13, 2016 (1 year ago)
Published:
September 26, 2017 at 09:20:18 CET
Submissions:
3
Open source:
No
Project page / code:
n/a
Reference:
C. Long, A. Haizhou, S. Chong, Z. Zijie, B. Bo. Online Multi-Object Tracking with Convolutional Neural Networks. In 2017 IEEE International Conference on Image Processing (ICIP), 2017.
Benchmark performance:
| MOTA | MOTP | FAF | MT | ML | FP | FN | ID Sw. | Frag | Specifications | Detector |
| 38.5 | 72.6 | 0.7 | 8.7 % | 37.4 % | 4,005 | 33,203 | 586 | 1,263 | GTX 1080, 2.3 GHz, 1 Core | Public |
| IDF1 | ID Precision | ID Recall |
| 47.1 | 68.4 | 35.9 |
Detailed performance:
| Sequence | MOTA | IDF1 | MOTP | FAF | GT | MT | ML | FP | FN | ID Sw | Frag |
| TUD-Crossing | 61.3 | 64.1 | 73.1 | 0.1 | 13 | 38.5 % | 15.4 % | 14 | 401 | 12 | 27 |
| PETS09-S2L2 | 38.9 | 34.3 | 70.8 | 1.3 | 42 | 2.4 % | 9.5 % | 552 | 5,164 | 179 | 328 |
| ETH-Jelmoli | 52.9 | 65.5 | 75.9 | 0.4 | 45 | 22.2 % | 28.9 % | 178 | 1,001 | 16 | 37 |
| ETH-Linthescher | 39.4 | 47.5 | 75.2 | 0.2 | 197 | 8.6 % | 62.9 % | 185 | 5,185 | 44 | 71 |
| ETH-Crossing | 34.2 | 42.3 | 79.5 | 0.1 | 26 | 3.8 % | 61.5 % | 20 | 630 | 10 | 13 |
| AVG-TownCentre | 28.4 | 44.7 | 66.9 | 2.1 | 226 | 4.0 % | 27.9 % | 941 | 4,005 | 169 | 412 |
| ADL-Rundle-1 | 35.3 | 43.6 | 72.2 | 1.1 | 32 | 12.5 % | 31.3 % | 573 | 5,413 | 39 | 81 |
| ADL-Rundle-3 | 43.3 | 51.9 | 77.6 | 1.0 | 44 | 15.9 % | 20.5 % | 643 | 5,076 | 47 | 85 |
| KITTI-16 | 40.7 | 61.1 | 68.9 | 0.8 | 17 | 17.6 % | 11.8 % | 160 | 831 | 18 | 35 |
| KITTI-19 | 34.3 | 50.6 | 67.7 | 0.6 | 62 | 3.2 % | 32.3 % | 663 | 2,805 | 43 | 147 |
| Venice-1 | 39.1 | 48.7 | 71.8 | 0.2 | 17 | 23.5 % | 41.2 % | 76 | 2,692 | 9 | 27 |
Raw data:
