Skip to content

[CVPR 2025] This is the official implementation of [Full-DoF Egomotion Estimation for Event Cameras Using Geometric Solvers]

License

Notifications You must be signed in to change notification settings

jizhaox/relpose-event

Repository files navigation

1. Overview

This repository provides solvers to estimate egomotion for event cameras. It contains solvers for angular and linear velocities [1]. The core optimization algorithms are implemented in C++. We provide compiled MEX files compatible with both Ubuntu 20.04 + Matlab 2018b, as well as Windows 10 + Matlab R2022a. The data generation and demonstration scripts are written in MATLAB.

[1] Ji Zhao, Banglei Guan, Zibin Liu, and Laurent Kneip. Full-DoF Egomotion Estimation for Event Cameras Using Geometric Solvers. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025.

@inproceedings{zhao2025full,
	title={Full-DoF Egomotion Estimation for Event Cameras Using Geometric Solvers},
	author={Zhao, Ji and Guan, Banglei and Liu, Zibin and Kneip, Laurent},
	booktitle={IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
	year={2025}
}

Main Contributors: Ji Zhao and Zibin Liu

2. Quick Start

run test_2dof_solver.m to estimate linear velocity only.

run test_5dof_solver_adam.m to esimtate angular and linear velocities.

If you want to compile the MEX file, change the path for EIGEN library in file compile.m and run it.

About

[CVPR 2025] This is the official implementation of [Full-DoF Egomotion Estimation for Event Cameras Using Geometric Solvers]

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published