Using Co-Occurrence and Segmentation to Learn Feature-Based Object Models from Video | IEEE Conference Publication | IEEE Xplore