Semantic Image Segmentation with Task-Specific Edge Detection Using CNNs and a Discriminatively Trained Domain Transform

被引:117
作者
Chen, Liang-Chieh [1 ]
Barron, Jonathan T. [1 ]
Papandreou, George [1 ]
Murphy, Kevin [1 ]
Yuille, Alan L. [1 ]
机构
[1] Google Inc, Mountain View, CA 94043 USA
来源
2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR) | 2016年
关键词
D O I
10.1109/CVPR.2016.492
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Deep convolutional neural networks (CNNs) are the backbone of state-of-art semantic image segmentation systems. Recent work has shown that complementing CNNs with fully-connected conditional random fields (CRFs) can significantly enhance their object localization accuracy, yet dense CRF inference is computationally expensive. We propose replacing the fully-connected CRF with domain transform (DT), a modern edge-preserving filtering method in which the amount of smoothing is controlled by a reference edge map. Domain transform filtering is several times faster than dense CRF inference and we show that it yields comparable semantic segmentation results, accurately capturing object boundaries. Importantly, our formulation allows learning the reference edge map from intermediate CNN features instead of using the image gradient magnitude as in standard DT filtering. This produces task-specific edges in an end-to-end trainable system optimizing the target semantic segmentation quality.
引用
收藏
页码:4545 / 4554
页数:10
相关论文
共 46 条
[1]  
[Anonymous], 2015, ARXIV150401013
[2]  
[Anonymous], 2015, CVPR
[3]  
[Anonymous], 2011, ICCV
[4]  
[Anonymous], 2014, P SSST EMNLP 2014 8
[5]  
[Anonymous], 2014, IJCV
[6]  
[Anonymous], 2016, ICLR
[7]  
[Anonymous], 2013, PAMI
[8]  
[Anonymous], 2015, CVPR
[9]  
[Anonymous], 2015, ICLR
[10]  
[Anonymous], 2015, ICCV