Table 6 Analysis of \(\alpha\), \(\beta\) and \(\gamma\) Loss weights on MOT16 train datasets.
From: Dual attention for multi object tracking with intra sample context and cross sample interaction
Loss weights | MOTA \(\uparrow\) | IDF1 \(\uparrow\) | MT \(\uparrow\) | ML \(\downarrow\) | IDs \(\downarrow\) |
|---|---|---|---|---|---|
[0.5, 0.5, 0.5] | 61.9 | 63.6 | 195 | 79 | 1303 |
[0.5, 0.5, 1.0] | 57.9 | 60.3 | 171 | 94 | 1419 |
[0.5, 1.0, 0.5] | 60.9 | 61.6 | 194 | 76 | 1486 |
[0.5, 1.0, 1.0] | 60.2 | 61.2 | 188 | 90 | 1477 |
[1.0, 0.5, 0.5] | 61.5 | 61.9 | 194 | 87 | 1469 |
[1.0, 0.5, 1.0] | 59.5 | 59.8 | 178 | 88 | 1456 |
[1.0, 1.0, 0.5] | 61.7 | 61.3 | 199 | 72 | 1441 |
[1.0, 1.0, 1.0] | 61.1 | 61.4 | 183 | 85 | 1395 |
