This paper strives to classify and detect the relationship between object tubelets appearing within a video as a <subject-predicate-object> triplet. Where existing works treat object proposals or tubelets as single entities and model their relations a posteriori, we propose to classify and detect predicates for pairs of object tubelets a priori. We also propose Social Fabric: an encoding that represents a pair of object tubelets as a composition of interaction primitives. These primitives are learned over all relations, resulting in a compact representation able to localize and classify relations from the pool of co-occurring object tubelets across all timespans in a video. The encoding enables our two-stage network. In the first stage, we train Social Fabric to suggest proposals that are likely interacting. We use the Social Fabric in the second stage to simultaneously fine-tune and predict predicate labels for the tubelets. Experiments demonstrate the benefit of early video relation modeling, our encoding and the two-stage architecture, leading to a new state-of-the-art on two benchmarks. We also show how the encoding enables query-by-primitive-example to search for spatio-temporal video relations. Code: https://github.com/shanshuo/Social-Fabric.
@article{arxiv.2108.08363,
title = {Social Fabric: Tubelet Compositions for Video Relation Detection},
author = {Shuo Chen and Zenglin Shi and Pascal Mettes and Cees G. M. Snoek},
journal= {arXiv preprint arXiv:2108.08363},
year = {2021}
}