Zhengxin Zhang, Qingjie Liu, Yunhong Wang
Road extraction from aerial images has been a hot research topic in the field of remote sensing image analysis. In this letter, a semantic segmentation neural network which combines the strengths of residual learning and U-Net is proposed for road area extraction. The network is built with residual units and has similar architecture to that of U-Net. The benefits of this model is two-fold: first, residual units ease training of deep networks. Second, the rich skip connections within the network could facilitate information propagation, allowing us to design networks with fewer parameters however better performance. We test our network on a public road dataset and compare it with U-Net and other two state of the art deep learning based road extraction methods. The proposed approach outperforms all the comparing methods, which demonstrates its superiority over recently developed state of the arts.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Medical Image Segmentation | Anatomical Tracings of Lesions After Stroke (ATLAS) | Dice | 0.4702 | ResUNet |
| Medical Image Segmentation | Anatomical Tracings of Lesions After Stroke (ATLAS) | IoU | 0.3549 | ResUNet |
| Medical Image Segmentation | Anatomical Tracings of Lesions After Stroke (ATLAS) | Precision | 0.5941 | ResUNet |
| Medical Image Segmentation | Anatomical Tracings of Lesions After Stroke (ATLAS) | Recall | 0.4537 | ResUNet |
| Medical Image Segmentation | ROSE-2 | Dice Score | 67.25 | ResU-Net |
| Medical Image Segmentation | CHASE_DB1 | AUC | 0.9779 | Residual U-Net |
| Medical Image Segmentation | CHASE_DB1 | F1 score | 0.78 | Residual U-Net |
| Medical Image Segmentation | ROSE-1 SVC-DVC | Dice Score | 74.61 | ResU-Net |
| Medical Image Segmentation | ROSE-1 SVC | Dice Score | 74.61 | ResU-Net |
| Medical Image Segmentation | STARE | F1 score | 0.8388 | Residual U-Net |
| Medical Image Segmentation | ROSE-1 DVC | Dice Score | 65.67 | ResU-Net |
| Medical Image Segmentation | DRIVE | AUC | 0.9779 | Residual U-Net |
| Medical Image Segmentation | DRIVE | F1 score | 0.8149 | Residual U-Net |
| Medical Image Segmentation | LUNA | AUC | 0.9849 | Residual U-Net |
| Medical Image Segmentation | LUNA | F1 score | 0.969 | Residual U-Net |
| Medical Image Segmentation | Kaggle Skin Lesion Segmentation | AUC | 0.9396 | Residual U-Net |
| Medical Image Segmentation | Kaggle Skin Lesion Segmentation | F1 score | 0.8799 | Residual U-Net |
| Semantic Segmentation | BJRoad | IoU | 54.24 | Res-UNet |
| 10-shot image generation | BJRoad | IoU | 54.24 | Res-UNet |
| Retinal Vessel Segmentation | ROSE-2 | Dice Score | 67.25 | ResU-Net |
| Retinal Vessel Segmentation | CHASE_DB1 | AUC | 0.9779 | Residual U-Net |
| Retinal Vessel Segmentation | CHASE_DB1 | F1 score | 0.78 | Residual U-Net |
| Retinal Vessel Segmentation | ROSE-1 SVC-DVC | Dice Score | 74.61 | ResU-Net |
| Retinal Vessel Segmentation | ROSE-1 SVC | Dice Score | 74.61 | ResU-Net |
| Retinal Vessel Segmentation | STARE | F1 score | 0.8388 | Residual U-Net |
| Retinal Vessel Segmentation | ROSE-1 DVC | Dice Score | 65.67 | ResU-Net |
| Retinal Vessel Segmentation | DRIVE | AUC | 0.9779 | Residual U-Net |
| Retinal Vessel Segmentation | DRIVE | F1 score | 0.8149 | Residual U-Net |