Semantic Segmentation with Modified Deep Residual Networks

摘要：

　　A novel semantic segmentation method is proposed,which consists of the following three parts:(Ⅰ)First,a simple yet effective data augmentation method is introduced without any extra GPU memory cost during training.(Ⅱ)Second,a deeper residual network is constructed through three effective techniques: dilated convolution,LSTM network and multi-scale prediction.(Ⅲ)Third,an online hard pixels mining is adopted to improve the segmentation performance.We combine these three parts to train an end-to-end network and achieve a new state-ofthe-art segmentation accuracy of 79.3%on PASCAL VOC 2012 test set at the time of submission.

关键词： Semantic segmentation Data augmentation Residual networks LSTM Multi-scale prediction

作者: Xinze Chen Guangliang Cheng Yinghao Cai Dayong Wen Heping Li

作者单位: NLPR,Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China

会议类型: 国际会议

会议名称: 第七届全国模式识别学术会议(The 7th Chinese Conference on Pattern Recognition,CCPR2016)

会议地点: 成都

会议语种:英文

页码: 42-54

在线出版日期: 2016-11-03（万方平台首次上网日期，不代表论文的发表时间）