Unifying Global and Local Scene Entities Modelling for Precise Action Spotting | IEEE Conference Publication | IEEE Xplore