Abstract

Image co-segmentation is popular with its ability to detour supervisory data by exploiting the common information in multiple im- ages. In this paper, we aim at a more challenging branch called scene image co-segmentation, which jointly segments multiple images captured from the same scene into regions corresponding to their respective classes. We first put forward a novel representation named Visual Relation Net- work (VRN) to organize multiple segments, and then search for mean- ingful segments for every image through voting on the network. Scalable topic-level random walk is then used to solve the voting problem. Ex- periments on the benchmark MSRC-v2, the more difficult LabelMe and SUN datasets show the superiority over the state-of-the-art methods.