Intra- and Inter-Reasoning Graph Convolutional Network for Saliency Prediction on 360° Images

Cubic projection can be utilized to divide 360° images into multiple rectilinear images, with little distortion. However, the existing saliency prediction models fail integrate semantic information of these images. In this paper, we address by proposing an intra- and inter-reasoning graph convolutional network for on (SalReGCN360). The whole framework contains six sub-networks, each which two branches. training phase, after utilizing Multiple Projection (MCP), are simultaneously put corresponding sub-networks. one branches, global features a single image extracted intra-graph inference module finely predict local other branch, contextual inter-graph effectively Finally, feature maps generated branches fusion, predicted. Extensive experiments popular datasets illustrate superiority proposed model, especially improvement in KLD metric.