Waymo Open Dataset is a collection of machine perception datasets and evaluation code for autonomous driving research.
The dataset addresses the need for large-scale, high-quality labeled data to advance autonomous driving technology. It provides three complementary datasets: the Perception dataset with high-resolution sensor data and labels for various perception tasks, the Motion dataset containing object trajectories and 3D maps across thousands of scenes, and the End-To-End Driving dataset with camera data paired with high-level driving commands. The codebase includes evaluation metrics and helper functions in TensorFlow to support model development across these different task domains.
Researchers working on autonomous driving perception, motion prediction, or end-to-end driving should consider this dataset. It is suited for projects requiring diverse sensor modalities, large-scale labeled examples, and standardized evaluation protocols. The dataset is particularly valuable for teams that need to benchmark against established metrics rather than define their own evaluation methodology.
Development activity shows consistent maintenance with a detailed release history documenting changes over time. The project provides helper functions and evaluation code alongside the dataset itself, indicating ongoing investment in making the resource accessible to researchers. The codebase is organized with clear separation between general-use code and limited-use code with specific licensing terms, suggesting careful attention to licensing compliance and use-case restrictions.
