Active learning with effective scoring functions for semi-supervised temporal action localization

Ding Li, Xuebing Yang, Yongqiang Tang, Chenyang Zhang, Wensheng Zhang, Lizhuang Ma

Abstract: Highlights•Temporal action localization requires large label cost.•The informativeness of different video samples varies in dataset.•Active learning conduces to the reduction of label cost.•Reasonable scoring functions are conducive to in formativeness estimation.