Deepen Refinery Expands Data Marketplace with ROVR’s Multimodal Driving Data

Deepen Refinery, a company building the verified data infrastructure for Physical AI, has expanded its Physical AI data marketplace with the addition of ROVR’s extensive multimodal driving data. This move significantly bolsters Deepen Refinery’s catalog, providing developers with a more comprehensive and diverse set of real-world sensor data.

ROVR, a provider of high-definition 3D data and Spatial AI, has amassed over 90,000 hours of multimodal driving data. This dataset includes high-precision LiDAR and camera data, along with IMU and GNSS data, all meticulously calibrated. The data spans a wide range of driving environments across the Americas, Europe, and Asia, offering a rich and diverse dataset for developers.

By integrating ROVR’s data into its platform, Deepen Refinery enhances the accessibility and usability of this valuable dataset. The company’s data infrastructure ensures that datasets are catalogued, structured, enriched, and validated, making them easier to discover and evaluate. This process maintains detailed information about the data’s provenance and characteristics, providing developers with a clear understanding of the data they are working with.

He further explained that ROVR’s large-scale real-world data collection capabilities, combined with Deepen Refinery’s data infrastructure, create a powerful resource for teams building autonomous and intelligent systems.

The addition of ROVR’s dataset to Deepen Refinery’s platform addresses a critical challenge in the Physical AI development process. Developers no longer need to spend significant time searching for, evaluating, and preparing disparate datasets. Instead, they can focus on model development, leveraging the high-quality and diverse data provided by ROVR.

For ROVR, this partnership provides a channel to make its high-quality datasets discoverable to a broader ecosystem of AI developers and organizations.

Deepen Refinery’s platform is designed to serve as a neutral data layer, connecting data sources with organizations building Physical AI. The platform brings together data from specialized contributors and makes it easier for AI teams to identify datasets based on factors such as modality, geography, environment, and application. Data can then be prepared and delivered in formats aligned with modern AI and robotics workflows, including ROS2, MCAP, HDF5, nuScenes, and LeRobot.

Refinery also provides a foundation for data validation and quality assessment, helping teams understand the characteristics and suitability of datasets before incorporating them into their development pipelines. This creates value on both sides of the marketplace: for data contributors, it offers a way to monetize real-world data without having to build and operate their own distribution infrastructure; for AI developers, it provides a centralized source for discovering and accessing data from multiple contributors, with greater visibility into provenance, characteristics, and quality.

In conclusion, the addition of ROVR’s 90,000+ hours of multimodal driving data to Deepen Refinery’s Physical AI data marketplace represents a significant step forward in the development of autonomous and intelligent systems. This collaboration not only enhances the quality and diversity of available data but also streamlines the development process, making it more efficient and effective for developers.