D-LinkNet: LinkNet with Pretrained Encoder and Dilated Convolution for High Resolution Satellite Imagery Road Extraction

Benchmark Model Rank Results
road-segementation-on-deepglobeD-LinkNet#2IoU: 0.6412
semantic-segmentation-on-bjroadD-LinkNet#4IoU: 57.96
semantic-segmentation-on-portoD-LinkNet#4IoU: 70.20