Aeva has released AevaScenes, reportedly the world’s first open dataset featuring synchronized, multi-sensor frequency-modulated continuous wave (FMCW) 4D lidar and camera data with object velocity measurements, semantic segmentation, tracking and lane-line annotations.
Designed to accelerate research in autonomous vehicle perception and expand the adoption of FMCW lidar, AevaScenes supports innovation in object detection, semantic segmentation, motion forecasting, scene flow and trajectory estimation. The dataset is now available for academic and non-commercial use.
The high-fidelity FMCW lidar data provides researchers and developers with accurate and dense range sensing, capturing depth and velocity information in challenging driving environments.
Rich multimodal sensor fusion combines FMCW 4D lidar with high-resolution camera imagery; the dataset supports research across detection, segmentation, tracking, sensor calibration and novel perception tasks.
Aeva’s interactive sensor diagram showcases wide and narrow fields of view for lidar and camera systems, enabling users to explore sensor characteristics and choose configurations that best suit their research needs
The ultra-long-range annotations provide what is reportedly the world’s first dataset with ultra-long-range annotations for object detection, semantic segmentation and lane detection at distances up to 400m.
“AevaScenes is the first dataset to bring together long-range FMCW lidar with velocity information and rich camera data, creating a new benchmark for perception research,” said James Reuther, chief engineer at Aeva. “By opening access to this level of fidelity and scale, we’re giving researchers the tools to push the boundaries of what’s possible in autonomous driving – whether that’s advancing detection and tracking or unlocking entirely new approaches to motion understanding.”
Key features
AevaScenes includes 100 curated sequences captured in and around the San Francisco Bay Area, covering urban and highway driving across day and night conditions. The dataset contains 10,000 frames of time-synchronized FMCW lidar and RGB camera data at 10Hz, collected using Aeva’s Mercedes Metris test vehicles.
The sensor suite consists of six FMCW lidar sensors (four wide FOV, two narrow FOV) and six high-resolution RGB cameras with matching FOVs, recording 4K RGB images at 10fps.
Data is provided in PCD point clouds, JPEG images and JSON annotations, totaling approximately 200GB (2GB per sequence). The dataset is evenly split between highway and urban environments and day and night conditions, with all sequences captured in clear weather on dry roads.
For more on open data sets, check out the April 2025 issue for an exclusive feature on how freely available datasets and scenario-sharing database platforms could speed AD development.