Short name:
TransCtr
Detector:
Public
Description:
TransCenter on public detections trained with COCO and MOT.
Reference:
Y. Xu, Y. Ban, G. Delorme, C. Gan, D. Rus, X. Alameda-Pineda. TransCenter: Transformers with Dense Queries for Multiple-Object Tracking. In arXiv, 2021.
Last submitted:
August 06, 2021 (3 years ago)
Published:
March 28, 2021 at 15:56:00 CET
Submissions:
4
Project page / code:
Open source:
Yes
Hardware:
RTX TITAN
Runtime:
1.0 Hz
Benchmark performance:
Sequence | MOTA | IDF1 | HOTA | MT | ML | FP | FN | Rcll | Prcn | AssA | DetA | AssRe | AssPr | DetRe | DetPr | LocA | FAF | ID Sw. | Frag |
MOT20 | 61.0 | 49.8 | 43.5 | 601 (48.4) | 192 (15.5) | 49,189 | 147,890 | 71.4 | 88.3 | 36.1 | 52.9 | 44.5 | 59.0 | 59.2 | 73.1 | 81.7 | 11.0 | 4,493 (0.0) | 8,950 (0.0) |
Detailed performance:
Sequence | MOTA | IDF1 | HOTA | MT | ML | FP | FN | Rcll | Prcn | AssA | DetA | AssRe | AssPr | DetRe | DetPr | LocA | FAF | ID Sw. | Frag |
MOT20-04 | 75.0 | 59.4 | 51.5 | 406 | 51 | 20,326 | 46,776 | 82.9 | 91.8 | 42.0 | 63.5 | 51.7 | 64.9 | 69.5 | 76.9 | 82.3 | 9.8 | 1,351 | 3,351 |
MOT20-06 | 45.6 | 35.6 | 30.7 | 89 | 72 | 14,441 | 55,970 | 57.8 | 84.2 | 23.0 | 41.5 | 29.3 | 45.5 | 46.9 | 68.3 | 80.6 | 14.3 | 1,834 | 3,210 |
MOT20-07 | 76.5 | 58.8 | 50.9 | 71 | 5 | 1,372 | 6,106 | 81.6 | 95.2 | 42.1 | 62.1 | 49.1 | 62.7 | 67.4 | 78.7 | 83.3 | 2.3 | 287 | 448 |
MOT20-08 | 31.5 | 31.6 | 26.5 | 35 | 64 | 13,050 | 39,038 | 49.6 | 74.7 | 21.2 | 33.7 | 26.8 | 46.0 | 40.0 | 60.1 | 79.3 | 16.2 | 1,021 | 1,941 |
Raw data: