M3ED: Multi-Robot Event Camera Data for High-Speed Motions

M3ED is the first multi-sensor event camera dataset focused on high-speed dynamic motions in robotics. It provides synchronized data from a car, legged robot, and UAV operating in challenging conditions like off-road trails and dense forests. The dataset includes high-resolution stereo event cameras, grayscale and RGB cameras, an IMU, a 64-beam LiDAR, and RTK localization, contributed by the Daniilidis Group and KumarRobotics.

Use Cases

• Developing event-based visual odometry algorithms based on high-speed stereo event camera data.
• Training multi-sensor fusion models for robot localization based on synchronized IMU, LiDAR, and RTK data.
• Benchmarking object detection and tracking in high-egomotion scenarios based on data from multiple independently moving objects.
• Evaluating perception robustness in challenging visual conditions like dense forests based on the described operational environments.

Strengths

• Includes synchronized data from three distinct robotic platforms: a car, a legged robot, and a UAV.
• Covers challenging operational scenarios such as off-road trails, dense forests, and aggressive flight maneuvers.
• Provides high-resolution stereo event camera data at 1280×720 resolution.
• Integrates multiple sensor modalities including a 64-beam LiDAR and high-quality IMU.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Daniilidis Group, KumarRobotics

Collection Method

Data gathered from multiple robotic platforms (car, legged robot, UAV) operating in various challenging environments.

Time Range

null

Freshness

null

Geography

null