From: RoadSegNet: a deep learning framework for autonomous urban road detection
S. no. | Metric | Network | ||
---|---|---|---|---|
ResNet50 | XceptionNet | MobileNet-V2 | ||
Training dataset | ||||
1 | Global accuracy | 97.52 | 97.34 | 97.80 |
2 | Mean accuracy | 89.83 | 93.80 | 98.70 |
3 | Mean IoU | 81.49 | 77.02 | 80.32 |
4 | Weighted IoU | 95.34 | 95.45 | 96.26 |
5 | Mean BF score | 0.8139 | 0.8008 | 0.8386 |
Testing dataset | ||||
1 | Global accuracy | 95.79 | 96.39 | 96.35 |
2 | Mean accuracy | 73.99 | 82.20 | 87.03 |
3 | Mean IoU | 69.02 | 72.99 | 75.28 |
4 | Weighted IoU | 92.15 | 93.63 | 93.59 |
5 | Mean BF score | 0.7572 | 0.7669 | 0.7885 |
Validation dataset | ||||
1 | Global accuracy | 96.62 | 97.14 | 97.13 |
2 | Mean accuracy | 86.92 | 94.25 | 95.67 |
3 | Mean IoU | 76.24 | 74.32 | 74.33 |
4 | Weighted IoU | 93.67 | 94.92 | 94.99 |
5 | Mean BF score | 0.7939 | 0.7979 | 0.8153 |