1 paper
Zixin Fan, Jiahong Lu, Changsheng Zheng +2
Existing Security-oriented Video Understanding (SVU) systems assume a \emph{closed world}, \ie static category sets, abundant labels, and the premise that all event types are known…