Universe Points Representation Learning for Partial Multi-Graph Matching
This work addresses a more general and challenging graph matching problem for applications like computer vision, but it appears incremental as it builds on recent deep learning progress for graphs.
The paper tackles the partial multi-graph matching problem with cycle consistency guarantees, proposing a data-driven method (URL) that uses an object-to-universe formulation and learns latent representations of abstract universe points, advancing state of the art in semantic keypoint matching on Pascal VOC, CUB, and Willow datasets and demonstrating scalability to large graphs and robustness to high partiality in synthetic experiments.
Many challenges from natural world can be formulated as a graph matching problem. Previous deep learning-based methods mainly consider a full two-graph matching setting. In this work, we study the more general partial matching problem with multi-graph cycle consistency guarantees. Building on a recent progress in deep learning on graphs, we propose a novel data-driven method (URL) for partial multi-graph matching, which uses an object-to-universe formulation and learns latent representations of abstract universe points. The proposed approach advances the state of the art in semantic keypoint matching problem, evaluated on Pascal VOC, CUB, and Willow datasets. Moreover, the set of controlled experiments on a synthetic graph matching dataset demonstrates the scalability of our method to graphs with large number of nodes and its robustness to high partiality.