Neighbourhood Consensus Networks

TL;DR

This paper introduces a neural network that leverages neighbourhood consensus patterns to improve dense correspondence matching between images, handling appearance differences and ambiguities without requiring explicit geometric models.

Contribution

It presents an end-to-end trainable CNN architecture that identifies spatially consistent matches using neighbourhood consensus, trained with weak supervision, and achieves state-of-the-art results on multiple benchmarks.

Findings

Achieves state-of-the-art performance on PF Pascal dataset.

Effective training with weak supervision from image pairs.

Applicable to both category- and instance-level matching.

Abstract

We address the problem of finding reliable dense correspondences between a pair of images. This is a challenging task due to strong appearance differences between the corresponding scene elements and ambiguities generated by repetitive patterns. The contributions of this work are threefold. First, inspired by the classic idea of disambiguating feature matches using semi-local constraints, we develop an end-to-end trainable convolutional neural network architecture that identifies sets of spatially consistent matches by analyzing neighbourhood consensus patterns in the 4D space of all possible correspondences between a pair of images without the need for a global geometric model. Second, we demonstrate that the model can be trained effectively from weak supervision in the form of matching and non-matching image pairs without the need for costly manual annotation of point to point correspondences. Third, we show the proposed neighbourhood consensus network can be applied to a range of matching tasks including both category- and instance-level matching, obtaining the state-of-the-art results on the PF Pascal dataset and the InLoc indoor visual localization benchmark.