Multi-label learning with missing labels for image annotation and facial action unit recognition
作者：
Highlights：
• An inductive method is proposed to handle missing labels in multi-label learning.
• The label bias of treating missing labels as negative labels is avoided.
• Label consistency, example-level and class-level smoothness are considered.
• We present an efficient algorithm to learn a parametric classifier.
• The proposed method is applied to image annotation and facial action recognition.
摘要
Highlights•An inductive method is proposed to handle missing labels in multi-label learning.•The label bias of treating missing labels as negative labels is avoided.•Label consistency, example-level and class-level smoothness are considered.•We present an efficient algorithm to learn a parametric classifier.•The proposed method is applied to image annotation and facial action recognition.
论文关键词：Multi-label learning,Missing labels,Image annotation,Facial action unit recognition
论文评审过程：Received 4 March 2014, Revised 29 November 2014, Accepted 26 January 2015, Available online 12 February 2015.
论文官网地址：https://doi.org/10.1016/j.patcog.2015.01.022
