Benchmark:
MOT17 |
Short name:
CTTrackPub
Detector:
Public
Description:
Our model takes the current frame, the previous frame, and a heatmap rendered from previous tracking results as input, and predicts the current detection heatmap as well as their offsets to centers in the previous frame.
Reference:
X. Zhou, V. Koltun, P. Kr"ahenb"uhl. Tracking Objects as Points. In ECCV, 2020.
Last submitted:
May 28, 2020 (4 years ago)
Published:
November 02, 2020 at 17:07:05 CET
Submissions:
2
Project page / code:
Open source:
No
Hardware:
4 GHZ, 12 Core, Titan Xp
Runtime:
17.0 Hz
Benchmark performance:
Sequence | MOTA | IDF1 | HOTA | MT | ML | FP | FN | Rcll | Prcn | AssA | DetA | AssRe | AssPr | DetRe | DetPr | LocA | FAF | ID Sw. | Frag |
MOT17 | 61.5 | 59.6 | 48.2 | 621 (26.4) | 752 (31.9) | 14,076 | 200,672 | 64.4 | 96.3 | 47.8 | 49.0 | 53.9 | 70.6 | 51.9 | 77.6 | 81.7 | 0.8 | 2,583 (40.1) | 4,965 (77.1) |
Detailed performance:
Sequence | MOTA | IDF1 | HOTA | MT | ML | FP | FN | Rcll | Prcn | AssA | DetA | AssRe | AssPr | DetRe | DetPr | LocA | FAF | ID Sw. | Frag |
MOT17-01-DPM | 41.6 | 44.2 | 36.9 | 5 | 8 | 496 | 3,252 | 49.6 | 86.6 | 37.6 | 36.5 | 41.9 | 65.0 | 39.9 | 69.7 | 79.6 | 1.1 | 22 | 58 |
MOT17-01-FRCNN | 41.0 | 42.1 | 35.5 | 6 | 9 | 571 | 3,207 | 50.3 | 85.0 | 34.6 | 36.8 | 39.3 | 63.7 | 40.6 | 68.6 | 79.5 | 1.3 | 25 | 61 |
MOT17-01-SDP | 41.8 | 44.3 | 36.8 | 7 | 8 | 612 | 3,112 | 51.8 | 84.5 | 36.5 | 37.4 | 41.2 | 64.6 | 41.5 | 67.8 | 79.4 | 1.4 | 27 | 65 |
MOT17-03-DPM | 79.3 | 71.6 | 56.9 | 94 | 8 | 1,142 | 20,297 | 80.6 | 98.7 | 52.7 | 61.8 | 58.1 | 74.4 | 65.0 | 79.5 | 81.9 | 0.8 | 191 | 525 |
MOT17-03-FRCNN | 79.6 | 72.7 | 57.5 | 93 | 7 | 1,234 | 19,945 | 80.9 | 98.6 | 53.7 | 62.0 | 58.8 | 74.8 | 65.2 | 79.4 | 81.9 | 0.8 | 180 | 508 |
MOT17-03-SDP | 80.0 | 72.0 | 57.5 | 93 | 8 | 1,223 | 19,530 | 81.3 | 98.6 | 53.5 | 62.2 | 58.7 | 75.0 | 65.5 | 79.3 | 81.9 | 0.8 | 181 | 526 |
MOT17-06-DPM | 54.8 | 42.0 | 35.9 | 54 | 63 | 314 | 4,839 | 58.9 | 95.7 | 28.9 | 45.2 | 48.8 | 41.8 | 47.8 | 77.7 | 82.2 | 0.3 | 175 | 244 |
MOT17-06-FRCNN | 55.6 | 42.9 | 36.7 | 57 | 59 | 363 | 4,676 | 60.3 | 95.1 | 29.5 | 46.0 | 48.3 | 42.9 | 48.9 | 77.1 | 82.0 | 0.3 | 190 | 264 |
MOT17-06-SDP | 55.5 | 43.8 | 37.1 | 56 | 61 | 354 | 4,712 | 60.0 | 95.2 | 30.2 | 45.9 | 49.1 | 43.5 | 48.7 | 77.3 | 82.0 | 0.3 | 181 | 262 |
MOT17-07-DPM | 44.8 | 42.0 | 33.6 | 11 | 16 | 1,322 | 7,851 | 53.5 | 87.2 | 30.4 | 38.1 | 34.5 | 63.3 | 42.1 | 68.7 | 79.9 | 2.6 | 147 | 275 |
MOT17-07-FRCNN | 45.5 | 41.5 | 33.4 | 13 | 15 | 1,263 | 7,785 | 53.9 | 87.8 | 29.8 | 38.5 | 33.7 | 63.7 | 42.4 | 69.1 | 79.8 | 2.5 | 156 | 289 |
MOT17-07-SDP | 45.2 | 42.4 | 33.9 | 13 | 15 | 1,332 | 7,775 | 54.0 | 87.3 | 30.6 | 38.3 | 34.6 | 63.8 | 42.4 | 68.6 | 79.8 | 2.7 | 147 | 279 |
MOT17-08-DPM | 26.5 | 32.2 | 30.5 | 11 | 37 | 378 | 15,066 | 28.7 | 94.1 | 39.5 | 23.8 | 46.3 | 68.9 | 24.5 | 80.5 | 84.5 | 0.6 | 88 | 146 |
MOT17-08-FRCNN | 26.5 | 31.9 | 30.5 | 11 | 36 | 332 | 15,113 | 28.5 | 94.8 | 39.5 | 23.7 | 46.3 | 69.2 | 24.3 | 81.1 | 84.6 | 0.5 | 89 | 141 |
MOT17-08-SDP | 26.6 | 32.3 | 30.6 | 11 | 36 | 350 | 15,067 | 28.7 | 94.5 | 39.6 | 23.8 | 46.4 | 69.7 | 24.5 | 80.8 | 84.5 | 0.6 | 91 | 147 |
MOT17-12-DPM | 46.1 | 53.1 | 44.6 | 16 | 45 | 207 | 4,434 | 48.8 | 95.3 | 51.0 | 39.2 | 59.1 | 70.6 | 41.2 | 80.3 | 84.7 | 0.2 | 30 | 50 |
MOT17-12-FRCNN | 46.1 | 52.6 | 44.5 | 15 | 45 | 197 | 4,443 | 48.7 | 95.5 | 50.6 | 39.2 | 59.0 | 70.1 | 41.1 | 80.5 | 84.7 | 0.2 | 30 | 48 |
MOT17-12-SDP | 46.0 | 53.0 | 44.6 | 16 | 45 | 221 | 4,426 | 48.9 | 95.0 | 50.9 | 39.2 | 59.0 | 70.6 | 41.2 | 80.1 | 84.7 | 0.2 | 30 | 52 |
MOT17-14-DPM | 31.6 | 36.6 | 26.9 | 13 | 78 | 636 | 11,812 | 36.1 | 91.3 | 28.1 | 26.0 | 31.9 | 57.7 | 27.5 | 69.7 | 78.5 | 0.8 | 196 | 331 |
MOT17-14-FRCNN | 31.6 | 37.6 | 27.6 | 13 | 77 | 780 | 11,653 | 37.0 | 89.8 | 29.0 | 26.5 | 32.7 | 58.9 | 28.2 | 68.5 | 78.3 | 1.0 | 202 | 350 |
MOT17-14-SDP | 31.7 | 37.1 | 27.3 | 13 | 76 | 749 | 11,677 | 36.8 | 90.1 | 28.5 | 26.5 | 32.3 | 57.8 | 28.2 | 69.0 | 78.4 | 1.0 | 205 | 344 |
Raw data: