neulky / EventBasedVisualOdometry

Implementation of Event-Based Visual Odometry algorithm - Group project for 16-720 Computer Vision / 16-833 SLAM

EVO

Matlab/C++ implementation of Event-Based Multi-View Stereo (mapping half of Event-Based Visual Odometry).

Prerequisites

Matlab: standalone - just need to get data.

ROS node (work in progress):

• OpenCV
• Eigen
• rpg_dvs_ros

Data

Get datasets from University of Zurich's [Event-Camera Dataset].

Matlab: download data in txt format, and run ExtractData on it to get a mat file.

ROS node: download rosbag to data folder and point emvs.launch to it.

Running

Matlab: Run EMVS.m

ROS node: roslaunch evo emvs.launch

Authors

• Max Hu
• Aum Jadhav
• Benjamin Kuo
• Cyrus Liu
• Maitreya Naik
• Kazu Otani

References

• Rebecq, Henri, Guillermo Gallego, and Davide Scaramuzza. ”EMVS: Event-based Multi-View Stereo.” British Machine Vision Conference (BMVC). No. EPFL-CONF-221504. 2016.

• Henri Rebecq, Timo Horstschaefer, Guillermo Gallego, Davide Scaramuzza, "EVO: A Geometric Approach to Event-based 6-DOF Parallel Tracking and Mapping in Real-time," IEEE Robotics and Automation Letters (RA-L), 2016.

• Gallup, David, et al. ”Real-time plane-sweeping stereo with multiple sweeping directions.” Computer Vision and Pattern Recognition, 2007. CVPR’07. IEEE Conference on. IEEE, 2007

• event-based_vision_resources from UZH group

• Image Reconstruction from an Event Camera - also from UZH