Dynamic Map Platform Co., Ltd. (Head Office: Shibuya-ku, Tokyo; CEO & President: Shuichi Yoshimura; “Dynamic Map Platform”) is pleased to announce that the AI-native sample dataset released on Hugging Face in June 2026 is now available through the Voxel51 Dataset Zoo, a widely used dataset catalog in the computer vision community.
Dynamic Map Platform’s AI-native dataset, which has accumulated more than 4,500 downloads on Hugging Face, has been listed in the Voxel51 Dataset Zoo, the dataset catalog maintained by Voxel51, the company behind the FiftyOne open-source project. As a result, researchers and developers can now discover and access Dynamic Map Platform’s datasets directly within their everyday Voxel51 workflows, further expanding opportunities for AI research and development.
- Dataset URL: https://docs.voxel51.com/dataset_zoo/datasets_hf/hard_intersection_multimodal_sample.html
■Background of the Voxel51 Dataset Zoo Listing
In recent years, real-world data has become increasingly important for training, evaluation, and validation in the development of Physical AI applications such as autonomous driving and robotics.
Leveraging its expertise in creating and providing high-precision 3D data for autonomous driving, Dynamic Map Platform is advancing its Data for AI business initiative. As part of this effort, Dynamic Map Platform released an AI-native sample dataset on Hugging Face in June 2026, integrating multiple data modalities including:
- Camera imagery
- LiDAR point clouds
- High-definition 3D map data
- Vehicle trajectory data
- Semantic annotations
- 3D Gaussian Splatting (3DGS) data
Designed as a multimodal dataset that integrates multiple real-world data sources, the dataset is intended to support research and development in autonomous driving AI and Physical AI. Since its release, it has been downloaded more than 4,500 times, and is now available through the Voxel51 Dataset Zoo, acurated catalog of datasets for computer vision and AI research.
■Expected Impact of the Voxel51 Dataset Zoo Listing
Voxel51 provides a unified platform for the physical AI data loop and maintains both the commercial platform and the open-source project, FiftyOne. Voxel51 is widely used by AI researchers and computer vision developers to visualize and analyze images, point cloud data, annotations, and other data assets for AI model development and evaluation.
Within the platform, the Voxel51 Dataset Zoo serves as a centralized catalog that enables researchers to easily discover and access publicly available datasets. It includes numerous datasets widely used in image recognition, autonomous driving, and computer vision research, allowing users to quickly obtain the data needed for their projects.
With Dynamic Map Platform’s AI-native dataset now included in the Voxel51 Dataset Zoo, users can directly load, visualize, analyze, and utilize the data for AI model development and research workflows. This significantly improves the discoverability and accessibility of Dynamic Map Platform’s datasets and is expected to drive broader adoption within the global research community.
Dynamic Map Platform will continue leveraging its expertise in real-world data collection and management developed through its autonomous driving data business. Beyond autonomous driving, Dynamic Map Platform aims to support research and development across a wide range of Physical AI domains, including robotics and logistics automation, by providing foundational data infrastructure that accelerates AI innovation and real-world deployment.
