Restricted set classification: Who is there?
作者：
Highlights：
• We define three new problems: who-is-who, who-is-missing and who-is-there.
• Related areas: object tracking, relaxation labelling, multi-instance classification.
• We propose a solution based on the Hungarian assignment algorithm.
• Our solution outperforms provably the original classifier and a greedy approach.
• Our real-life examples are naming the fish in a tank and labelling chess pieces.
摘要
Highlights•We define three new problems: who-is-who, who-is-missing and who-is-there.•Related areas: object tracking, relaxation labelling, multi-instance classification.•We propose a solution based on the Hungarian assignment algorithm.•Our solution outperforms provably the original classifier and a greedy approach.•Our real-life examples are naming the fish in a tank and labelling chess pieces.
论文关键词：Pattern recognition,Object classification,Restricted set classification,Compound decision problem,Chess pieces classification
论文评审过程：Received 31 March 2015, Revised 4 April 2016, Accepted 23 August 2016, Available online 30 September 2016, Version of Record 15 October 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2016.08.028
