OpenVL: Abstracting Vision Tasks Using a Segment-Based Language Model | IEEE Conference Publication | IEEE Xplore