Weakly-supervised temporal action localization aims to localize actioninstances temporal boundary and identify the corresponding action category with only video-level labels . Traditional methods mainly focus on foreground and background frames separation with only a single attention branch and classactivation sequence . We propose an action-context modeling network termed ACM-Net, which measures the likelihood of each temporal point being action instance, context, or non-action background, simultaneously . We conduct extensive experiments on two benchmarkdatasets, THUMOS-14 and ActivityNet-1.3.3 . The experiments show that our method outperforms current state-of-the-art methods, and even achieve comparableperformance with fully-supervisory methods . Code can be found athttps://://://github.com/ispc-lab/ACM-net/ACm-Net. To read more from IISC-Lab/ISICL:

Author(s) : Sanqing Qu, Guang Chen, Zhijun Li, Lijun Zhang, Fan Lu, Alois Knoll

Links : PDF - Abstract

Code :

Keywords : action - acm - temporal - net - methods -

Leave a Reply

Your email address will not be published. Required fields are marked *