Video visual relation detection. which aims to detect the visual relations between objects in the form of relation triplet <subject. predicate. object> (e. https://www.jmannino.com/