Table 2.
Comparison of segmentation results of several methods.
| Model | Backbone | mIoU (%) | Parameters (M) | FLOPs (G) | FPS |
|---|---|---|---|---|---|
| FCN | VGG16 | 75.0 | 32.75 | 89.8 | 14.73 |
| Unet | VGG16 | 79.2 | 43.93 | 184.4 | 27.76 |
| PSPNet | ResNet50 | 74.7 | 46.716 | 118.47 | 29.4 |
| DeepLabv3+ | Xception | 86.28 | 54.714 | 167.139 | 18.4 |
| SegNet | VGG16 | 77.9 | 29.44 | 126.34 | 9.45 |
| BiSeNet | Xception | 78.47 | 5.8 | 50.3 | 41.23 |
| ICNet | ResNet50 | 79.65 | 26.5 | 28.3 | 30.3 |
| Ours | MobileNetV2 | 85.57 | 2.89 | 15.326 | 42.89 |
Performance comparison of different models. Bold values indicate the best results.