A generic mid-level representation for semantic video analysis | IEEE Conference Publication | IEEE Xplore