A novel semantic segmentation method based on improved FCN for unmanned driving

Driverless semantic segmentation is one of the research hotspots in the field of computer vision in recent years. Therefore, this paper proposes an unmanned semantic segmentation method based on improved fully convolutional network. Based on the FCN model, this paper uses HRNet to replace the backbone network. HRNet adopts high-resolution feature fusion method to ensure the output of high-resolution feature maps. It is verified on the Cityscapes dataset. Through the comparison of ablation experiments, it is found that the proposed HRNet backbone network has an average increase of 21.54 % in mIoU compared with other backbone networks with FCN as the model, which proves the effectiveness of the proposed model. In addition, in order to prove the advanced nature of the method in this paper, four widely used models, UNet, PSPNet, DeepLabV3 and OCRNet, are selected for comparison. The average mIoU is increased by 11.77 %, which proves the advanced nature of the model in this paper.