Short name:
reID2track
Benchmark:
Description:
The algorithm is an re-idetification task that uses the faster R-CNN. The improvements are the increase in the resolution of the network and scaling of anchors. The network learns features for each object and these features using the faster RCNN algorithm and on top of that an identification is applied that creates features for each unique object for the purpose of re-identification.
The detection features from the conv-net are used for tracking. The algorithm uses cosine similarity for feature association. For tracking, simply the IDs the assigned based on the similarity threshold.
Hardware:
Intel(R) Xeon(R) CPU E5-2690v3 2.6GHz, 256GB RAM, Nvidia Quadro P6000
Detector:
Public
Processing:
Online
Last submitted:
May 07, 2018 (9 months ago)
Published:
May 07, 2018 at 11:41:19 CET
Submissions:
2
Open source:
No
Project page / code:
n/a
Reference:
Anonymous submission
Benchmark performance:
MOTA | MOTP | FAF | MT | ML | FP | FN | ID Sw. | Frag | Specifications | Detector |
44.6 | 76.9 | 1.3 | 15.8 % | 39.7 % | 22,451 | 284,213 | 6,134 | 13,786 | Intel(R) Xeon(R) CPU E5-2690v3 2.6GHz, 256GB RAM, Nvidia Quadro P6000 | Public |
IDF1 | ID Precision | ID Recall |
39.9 | 57.2 | 30.7 |
Detailed performance:
Sequence | MOTA | IDF1 | MOTP | FAF | GT | MT | ML | FP | FN | ID Sw | Frag |
MOT17-01-DPM | 20.4 | 26.7 | 73.2 | 0.2 | 24 | 4.2 % | 62.5 % | 105 | 4,990 | 37 | 177 |
MOT17-03-DPM | 37.5 | 26.2 | 76.1 | 2.4 | 148 | 9.5 % | 31.8 % | 3,583 | 59,666 | 2,129 | 3,184 |
MOT17-06-DPM | 38.6 | 35.5 | 74.4 | 0.3 | 222 | 8.1 % | 54.5 % | 316 | 6,819 | 100 | 359 |
MOT17-07-DPM | 26.9 | 30.8 | 74.4 | 1.3 | 60 | 6.7 % | 58.3 % | 647 | 11,519 | 179 | 588 |
MOT17-08-DPM | 20.5 | 25.2 | 79.6 | 1.2 | 76 | 7.9 % | 55.3 % | 742 | 15,931 | 131 | 286 |
MOT17-12-DPM | 35.3 | 43.9 | 77.2 | 0.5 | 91 | 8.8 % | 52.7 % | 406 | 5,166 | 38 | 148 |
MOT17-14-DPM | 14.4 | 19.2 | 75.2 | 0.3 | 164 | 2.4 % | 78.0 % | 205 | 15,518 | 100 | 457 |
MOT17-01-FRCNN | 27.3 | 41.5 | 76.5 | 3.0 | 24 | 25.0 % | 33.3 % | 1,367 | 3,292 | 33 | 76 |
MOT17-03-FRCNN | 56.8 | 47.9 | 78.0 | 1.1 | 148 | 24.3 % | 18.9 % | 1,576 | 43,298 | 376 | 1,071 |
MOT17-06-FRCNN | 54.6 | 43.8 | 78.6 | 0.4 | 222 | 22.1 % | 25.7 % | 445 | 4,769 | 141 | 345 |
MOT17-07-FRCNN | 32.8 | 37.4 | 74.8 | 2.7 | 60 | 10.0 % | 25.0 % | 1,359 | 9,740 | 251 | 564 |
MOT17-08-FRCNN | 22.2 | 31.5 | 80.1 | 1.2 | 76 | 9.2 % | 52.6 % | 766 | 15,583 | 75 | 185 |
MOT17-12-FRCNN | 36.6 | 49.2 | 78.3 | 0.7 | 91 | 13.2 % | 48.4 % | 601 | 4,863 | 34 | 130 |
MOT17-14-FRCNN | 18.4 | 30.6 | 71.3 | 3.6 | 164 | 6.7 % | 46.3 % | 2,693 | 11,998 | 398 | 796 |
MOT17-01-SDP | 38.5 | 40.8 | 73.6 | 2.5 | 24 | 29.2 % | 16.7 % | 1,105 | 2,725 | 134 | 284 |
MOT17-03-SDP | 72.1 | 52.6 | 77.6 | 0.8 | 148 | 45.3 % | 10.1 % | 1,272 | 27,060 | 830 | 2,396 |
MOT17-06-SDP | 56.7 | 46.0 | 76.5 | 0.5 | 222 | 29.7 % | 29.3 % | 622 | 4,309 | 170 | 332 |
MOT17-07-SDP | 44.7 | 45.5 | 75.5 | 2.1 | 60 | 18.3 % | 26.7 % | 1,027 | 8,043 | 267 | 694 |
MOT17-08-SDP | 28.5 | 29.7 | 78.2 | 1.1 | 76 | 13.2 % | 47.4 % | 664 | 14,211 | 231 | 517 |
MOT17-12-SDP | 40.8 | 55.1 | 78.4 | 0.9 | 91 | 20.9 % | 40.7 % | 805 | 4,275 | 47 | 198 |
MOT17-14-SDP | 29.6 | 32.5 | 72.2 | 2.9 | 164 | 6.1 % | 35.4 % | 2,145 | 10,438 | 433 | 999 |
Raw data:
n/a