Category-specific object segmentation via unsupervised discriminant shape
作者：
Highlights：
• A novel unsupervised discriminant shape is proposed for segmenting images.
• An unsupervised discriminant clustering method is built for precise shape prior.
• Our model can produce reliable segmentation for category-specific objects.
• Evaluations on 4 datasets demonstrate the effectiveness of our method.
摘要
Highlights•A novel unsupervised discriminant shape is proposed for segmenting images.•An unsupervised discriminant clustering method is built for precise shape prior.•Our model can produce reliable segmentation for category-specific objects.•Evaluations on 4 datasets demonstrate the effectiveness of our method.
论文关键词：Object segmentation,Unsupervised discriminant clustering,Graph cut
论文评审过程：Received 21 November 2015, Revised 11 November 2016, Accepted 12 November 2016, Available online 16 November 2016, Version of Record 24 November 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2016.11.009
