Table 2 Comparison experiment results on TuSimple. bold, italic, and bolditalic indicate the first, second, and third scores, respectively.
From: Aggregate global features into separable hierarchical lane detection transformer
Method | Accuracy (%) | FP (%) | FN (%) | PP | FPS |
|---|---|---|---|---|---|
SCNN34 | 94.77 | 0.0753 | 0.0737 | \(\checkmark\) | 21 |
RESA35 | 95.24 | 0.0685 | 0.0571 | – | 68 |
UFLD36 | 95.61 | 0.1918 | 0.0404 | – | 144 |
FastDraw37 | 95.20 | 0.0760 | 0.0450 | \(\checkmark\) | 90 |
ENet-SAD38 | 96.64 | 0.0602 | 0.0205 | \(\checkmark\) | 75 |
Line-CNN39 | 96.87 | 0.0442 | 0.0197 | \(\checkmark\) | 30 |
PINet40 | 96.70 | 0.0294 | 0.0442 | \(\checkmark\) | 30 |
PolyLaneNet41 | 93.36 | 0.0942 | 0.0933 | – | 115 |
LSTR27 | 96.18 | 0.0291 | 0.0338 | – | 100 |
Ours | 96.42 | 0.0279 | 0.0318 | – | 78 |
