Iterative Visual Reasoning Beyond Convolutions

Xinlei Chen, Li-Jia Li, Li Fei-Fei, Abhinav Gupta; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 7239-7248

We present a novel framework for iterative visual reasoning. Our framework goes beyond current recognition systems that lack the capability to reason beyond stack of convolutions. The framework consists of two core modules: a local module that uses spatial memory to store previous beliefs in parallel; and a global graph-reasoning module. Our graph has three components: a) a knowledge graph where we represent classes as nodes and build edges to encode different types of semantic relationships between them; b) a region graph of the current image where regions in the image are nodes and spatial relationships between these regions are edges; c) an assignment graph that assigns regions to class nodes. Both the local module and the global module roll-out iteratively and cross-feed predictions to each other to refine estimates. The final predictions are made by combining the best of both modules with an attention mechanism. We show strong performance over plain ConvNets, eg achieving an $8.4%$ absolute improvement on ADE measured by per-class average precision. Analysis also shows that the framework is resilient to missing regions for reasoning.