Dual-Resolution Correspondence Networks

TL;DR

DualRC-Net introduces a coarse-to-fine approach for dense pixel-wise image correspondence, leveraging dual-resolution features and a learnable refinement process to improve accuracy and efficiency, achieving state-of-the-art results.

Contribution

The paper presents DualRC-Net, a novel method that combines coarse and fine features with a learnable refinement to enhance pixel correspondence accuracy efficiently.

Findings

Achieves state-of-the-art results on HPatches, InLoc, and Aachen benchmarks.

Improves matching reliability and localization accuracy.

Reduces computational cost by avoiding expensive 4D convolutions on fine features.

Abstract

We tackle the problem of establishing dense pixel-wise correspondences between a pair of images. In this work, we introduce Dual-Resolution Correspondence Networks (DualRC-Net), to obtain pixel-wise correspondences in a coarse-to-fine manner. DualRC-Net extracts both coarse- and fine- resolution feature maps. The coarse maps are used to produce a full but coarse 4D correlation tensor, which is then refined by a learnable neighbourhood consensus module. The fine-resolution feature maps are used to obtain the final dense correspondences guided by the refined coarse 4D correlation tensor. The selected coarse-resolution matching scores allow the fine-resolution features to focus only on a limited number of possible matches with high confidence. In this way, DualRC-Net dramatically increases matching reliability and localisation accuracy, while avoiding to apply the expensive 4D convolution kernels on fine-resolution feature maps. We comprehensively evaluate our method on large-scale public benchmarks including HPatches, InLoc, and Aachen Day-Night. It achieves the state-of-the-art results on all of them.