Road Extraction by Deep Residual U-Net

29 Nov 2017  ·  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.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Lesion Segmentation Anatomical Tracings of Lesions After Stroke (ATLAS) ResUNet Dice 0.4702 # 2
IoU 0.3549 # 2
Precision 0.5941 # 2
Recall 0.4537 # 2
Retinal Vessel Segmentation CHASE_DB1 Residual U-Net F1 score 0.7800 # 12
AUC 0.9779 # 10
Retinal Vessel Segmentation DRIVE Residual U-Net F1 score 0.8149 # 13
AUC 0.9779 # 10
Skin Cancer Segmentation Kaggle Skin Lesion Segmentation Residual U-Net F1 score 0.8799 # 2
AUC 0.9396 # 2
Lung Nodule Segmentation LUNA Residual U-Net F1 score 0.9690 # 2
AUC 0.9849 # 2
Retinal Vessel Segmentation ROSE-1 DVC ResU-Net Dice Score 65.67 # 3
Retinal Vessel Segmentation ROSE-1 SVC ResU-Net Dice Score 74.61 # 4
Retinal Vessel Segmentation ROSE-1 SVC-DVC ResU-Net Dice Score 74.61 # 3
Retinal Vessel Segmentation ROSE-2 ResU-Net Dice Score 67.25 # 4
Retinal Vessel Segmentation STARE Residual U-Net F1 score 0.8388 # 3

Methods