ETO:Efficient Transformer-based Local Feature Matching by Organizing Multiple Homography Hypotheses

TL;DR

This paper introduces an efficient transformer-based local feature matching method that constructs multiple homography hypotheses and uses uni-directional cross-attention, achieving high accuracy with four times faster inference than existing methods.

Contribution

The proposed architecture combines multiple homography hypotheses with uni-directional cross-attention to improve efficiency and accuracy in local feature matching.

Findings

Achieves competitive accuracy with LoFTR on YFCC100M.

Boosts inference speed to 4 times faster than state-of-the-art transformer methods.

Demonstrates effectiveness across multiple datasets like Megadepth, ScanNet, and HPatches.

Abstract

We tackle the efficiency problem of learning local feature matching. Recent advancements have given rise to purely CNN-based and transformer-based approaches, each augmented with deep learning techniques. While CNN-based methods often excel in matching speed, transformer-based methods tend to provide more accurate matches. We propose an efficient transformer-based network architecture for local feature matching. This technique is built on constructing multiple homography hypotheses to approximate the continuous correspondence in the real world and uni-directional cross-attention to accelerate the refinement. On the YFCC100M dataset, our matching accuracy is competitive with LoFTR, a state-of-the-art transformer-based architecture, while the inference speed is boosted to 4 times, even outperforming the CNN-based methods. Comprehensive evaluations on other open datasets such as Megadepth, ScanNet, and HPatches demonstrate our method's efficacy, highlighting its potential to significantly enhance a wide array of downstream applications.