Table 1.
Detection accuracy and speed of all lane detection algorithms on KITTI and Caltech datasets.
| Algorithm | KITTI | Caltech | ||
|---|---|---|---|---|
| mAP (%) |
Speed (ms) |
mAP (%) |
Speed (ms) |
|
| Fast RCNN [25] | 49.87 | 2271 | 53.13 | 2140 |
| Faster RCNN [26] | 58.78 | 122 | 61.73 | 149 |
| Sliding window & CNN [23] | 68.98 | 79,000 | 71.26 | 42,000 |
| SSD [28] | 75.73 | 29.3 | 77.39 | 25.6 |
| Context & RCNN [45] | 79.26 | 197 | 81.75 | 136 |
| Yolo v1 (S × S) [27] | 72.21 | 44.7 | 73.92 | 45.2 |
| T-S Yolo v1 (S × 2S) | 74.67 | 45.1 | 75.69 | 45.4 |
| Yolo v2 (S × S) [29] | 81.64 | 59.1 | 82.81 | 58.5 |
| T-S Yolo v2 (S × 2S) | 83.16 | 59.6 | 84.07 | 59.2 |
| Yolo v3 (S × S) [31] | 87.42 | 24.8 | 88.44 | 24.3 |
| T-S Yolo v3 (S × 2S) | 88.39 | 25.2 | 89.32 | 24.7 |