SAGTA: Semi-automatic Ground Truth Annotation in crowd scenes | IEEE Conference Publication | IEEE Xplore