Research Article
SKT-MOT and DyTracker: A Multiobject Tracking Dataset and a Dynamic Tracker for Speed Skating Video
Table 4
Comparison with state-of-the-art MOT methods on the SKT-MOT dataset.
| Method | Ref. | MOTA | IDF1 | MOTP | FP | FN | IDSW | FPS |
| SST [7] | TPAMI2019 | 76.02 | 56.48 | 85.87 | 3,448 | 3,991 | 1427 | 4.2 | JDE [9] | ECCV2020 | 79.87 | 72.91 | 78.31 | 2,983 | 4,943 | 812 | 13.2 | LocalSORT [47] | JSS2021 | 83.14 | 76.82 | 82.43 | 2,365 | 4,352 | 624 | 11.8 | FairMOT [10] | IJCV2021 | 86.53 | 76.51 | 85.71 | 1,359 | 4,308 | 557 | 16.4 | ByteTrack [23] | ECCV2022 | 84.56 | 73.72 | 87.32 | 3,565 | 3,219 | 397 | 17.8 | DeepSORT [6] | ICIP2017 | 81.78 | 77.61 | 81.24 | 2,955 | 4,363 | 735 | 9.3 | DyTracker | This study | 93.70 | 92.39 | 87.79 | 939 | 3,670 | 154 | 8.9 |
|
|
TBD methods use the same detection results. The best results are in bold.
|