Waymo Open Dataset
Project record
Autonomous driving data collected by Waymo vehicles, released for research in August 2019 and since expanded into three datasets: Perception (lidar and camera data with labels for 2,030 segments), Motion (object trajectories and 3D maps for 103,354 segments), and End-to-End Driving (360-degree camera images and routing instructions for 5,000 segments). A companion repository provides the data format definitions, evaluation metrics, and TensorFlow helper code.257
- Website: Waymo Open Dataset website (external site: waymo.com)
- Documentation: About the dataset (external site: waymo.com)
- License: Waymo Dataset License Agreement for Non-Commercial Use (external site: waymo.com)
- Repository: waymo-open-dataset code repository (external site: github.com)
- Paper: Perception dataset paper (arXiv 1912.04838) (external site: arxiv.org)
- Paper: Motion dataset paper (arXiv 2104.10133) (external site: arxiv.org)
Availability and license
Overall availability
Free of charge, but downloads require signing in with a Google account and registering at waymo.com/open, and use is limited to non-commercial purposes under Waymo's custom license. Waymo's FAQ states that the license should not be considered an open source license.143
Availability is separate from permission: read the license before using or redistributing.
Waymo Dataset License Agreement for Non-Commercial Use (March 2025) (external site: waymo.com)4
Apache License 2.0 (repository code outside src/waymo_open_dataset/wdl_limited) (external site: raw.githubusercontent.com)65
BSD 3-clause copyright license plus a limited patent license (wdl_limited folders) (external site: raw.githubusercontent.com)65
The data license grants a personal, royalty-free license for research, teaching, scientific publication, personal experimentation, and benchmarking for academic or applied research publication only. It bars using the data, or models trained on it, to operate or assist a vehicle, in production systems, or for other primarily commercial purposes. Apart from small extracts used as illustrations in publications, the dataset may be redistributed only to others who have registered and accepted the terms; trained models may be published only with notices that carry the non-commercial terms downstream; users must include a set attribution and accept a limited non-assert toward Waymo. Code in each wdl_limited subfolder also carries a limited patent license that covers only the use case that license sets out in connection with the dataset, in compliance with the data license. The repository README's metadata still names an August 2019 version of the license, while the terms page shows the March 2025 version.45
Public materials checklist
| Item | Status | Notes and evidence |
|---|---|---|
| AccessCan the data be obtained, and on what terms? | Partial | Downloadable at no cost after Google sign-in and registration; the download and data pages redirect to Google sign-in. Use is restricted to non-commercial purposes.14 |
| ProvenanceAre the data's origins documented? | Public | Waymo states the data was collected by its vehicles, lists the sensors used for each dataset, and names collection locations (the Motion dataset's map locations include San Francisco, Phoenix, Mountain View, Los Angeles, Detroit, and Seattle). It notes the data mixes manual and autonomous driving without labeling which is which.237 |
| DocumentationIs there a datasheet, card, or equivalent documentation? | Public | The dataset site describes each component and notes its latest updates; the repository holds the format definitions, labeling specification, metrics code, tutorials, and a release history, and papers describe the Perception and Motion datasets.23589 |
| LicensingAre the licensing terms stated? | Public | Custom non-commercial data license published on the dataset site; repository code mainly Apache 2.0.46 |
| Stated limitationsDoes the documentation state known limitations or risks? | Public | Waymo states the data should be used only for research and not to evaluate the real-world performance of any vehicle, because manual and autonomous driving data are mixed, and that it is only a fraction of the data used to train the Waymo Driver. Images blur faces and license plates.23 |
What it is useful for
Organization context
Provenance and derivatives
Collected by Waymo's own vehicles. The August 2019 first release covered 1,000 driving segments recorded in Phoenix, Kirkland, Mountain View, and San Francisco; later versions added the Motion and End-to-End Driving datasets. Motion dataset 3D boxes were produced by a model trained on the Perception dataset.72
U.S. eligibility
Sources
This listing is not an endorsement, a safety assessment, or a federal approval.