U.S.-based sensor technology company Aeva has released AevaScenes, an open-access dataset that combines frequency modulated continuous wave (FMCW) 4D LiDAR with synchronized camera data, aiming to support research in autonomous vehicle perception.
The dataset, which the company says is the first of its kind, provides long-range sensor data with precise object velocity measurements and detailed annotations, including semantic segmentation, object tracking and lane line information. Aeva said AevaScenes is intended to help researchers advance object detection, motion forecasting and scene flow analysis in self-driving systems.
See also: LG Innotek Partners with Aeva to Develop 4D LiDAR for Autonomous and Industrial Applications
Captured in and around the San Francisco Bay Area using Aeva’s Mercedes Metris test vehicles, the dataset features 100 curated sequences with 10,000 synchronized frames of LiDAR and camera data. It includes six FMCW LiDAR sensors and six high-resolution 4K RGB cameras, covering both urban and highway settings in day and night conditions.
James Reuther, Aeva’s chief engineer, said the resource combines long-range LiDAR with velocity information and high-resolution imagery to set a new benchmark for perception research. The dataset, which totals about 200 GB, is available in standard formats such as PCD point clouds, JPEG images and JSON annotations.