1 paper
Elahe Vahdani, Yingli Tian
This paper addresses the challenge of point-supervised temporal action detection, in which only one frame per action instance is annotated in the training set. Self-training aims t…