Benchmark:
Short name:
TrackRCNN
Description:
TrackR-CNN baseline method, extends Mask R-CNN with 3D convolutions and association head. Detections come directly from TrackR-CNN.
Reference:
P. Voigtlaender, M. Krause, A. O\usep, J. Luiten, B. Sekar, A. Geiger, B. Leibe. MOTS: Multi-Object Tracking and Segmentation. In CVPR, 2019.
Last submitted:
April 08, 2020 (4 years ago)
Published:
April 08, 2020 at 16:31:58 CET
Submissions:
1
Project page / code:
Open source:
No
Hardware:
GTX 1080 TI
Runtime:
2.0 Hz
Benchmark performance:
| Sequence | sMOTSA | IDF1 | MOTSA | MOTSP | MODSA | MT | ML | TP | FP | FN | Rcll | Prcn | ID Sw. | Frag |
| CVPR 2020 MOTS Challenge | 40.6 | 42.4 | 55.2 | 76.1 | 56.9 | 127 (38.7) | 71 (21.6) | 19,628 | 1,261 | 12,641 | 60.8 | 94.0 | 567 (932.2) | 868 (1,427.0) |
Detailed performance:
| Sequence | sMOTSA | IDF1 | MOTSA | MOTSP | MODSA | MT | ML | TP | FP | FN | Rcll | Prcn | ID Sw. | Frag |
| MOTS20-01 | 37.5 | 51.3 | 57.6 | 69.8 | 58.8 | 6 | 1 | 2,065 | 239 | 1,041 | 66.5 | 89.6 | 38 | 76 |
| MOTS20-06 | 57.1 | 37.8 | 73.7 | 79.4 | 76.8 | 98 | 23 | 7,908 | 369 | 1,906 | 80.6 | 95.5 | 310 | 388 |
| MOTS20-07 | 22.4 | 35.3 | 34.5 | 69.4 | 35.9 | 4 | 28 | 5,063 | 438 | 7,815 | 39.3 | 92.0 | 184 | 294 |
| MOTS20-12 | 53.4 | 57.3 | 67.1 | 80.6 | 67.6 | 19 | 19 | 4,592 | 215 | 1,879 | 71.0 | 95.5 | 35 | 110 |
Raw data: