Leveraging temporal, contextual and ordering constraints for recognizing complex activities in video | IEEE Conference Publication | IEEE Xplore