开放集识别研究综述

高菲, 杨柳, 李晖

A survey on open set recognition

Fei Gao, Liu Yang, Hui Li

图1 传统分类器与开放集识别对比：(a)数据集的原始分布，包括五个KKCs和两个UUCs；(b)传统分类器的决策边界，其中UUCs会分到KKCs中；(c)开放集识别，能识别UUCs并单独分类

Fig.1 Comparison of traditional classification and open set recognition：(a) the distribution of original dataset including KKCs and UUCs，(b) the decision boundary of each KKC obtained by traditional classification methods with UUCs included in KKCs' spaces，(c) open set recognition which can recognize UUCs as a class