Learning to segment with image-level annotations

作者：

Highlights：

• Localization map generation is proposed by using the hypothesis-based classification.

• A novel multi-label loss is proposed to train the network based on localization maps.

• An effective method is proposed to predict the rough mask of the given training image.

• Our methods achieve new state-of-the-art results on PASCAL VOC 2012 benchmark.

摘要

Highlights•Localization map generation is proposed by using the hypothesis-based classification.•A novel multi-label loss is proposed to train the network based on localization maps.•An effective method is proposed to predict the rough mask of the given training image.•Our methods achieve new state-of-the-art results on PASCAL VOC 2012 benchmark.

论文关键词：Semantic segmentation,Weakly supervised,Deep learning

论文评审过程：Received 19 July 2015, Revised 17 January 2016, Accepted 18 January 2016, Available online 27 January 2016, Version of Record 23 August 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.01.015