Argoverse introduces two large-scale datasets for autonomous driving research: the Argoverse 3D Tracking dataset and the Argoverse Motion Forecasting dataset. These datasets are designed to support machine learning tasks such as 3D tracking and motion forecasting. The 3D Tracking dataset includes LiDAR measurements, 360° RGB video, front-facing stereo images, and 6-DOF localization, all aligned with detailed maps containing lane centerlines, drivable regions, and ground height. The Motion Forecasting dataset includes over 300,000 5-second tracked scenarios with identified vehicles for trajectory forecasting. Argoverse is the first dataset to include "HD maps" with 290 km of mapped lanes and geometric and semantic metadata. The data is released under a Creative Commons license, and the authors provide an API to connect map data with sensor information. The paper evaluates baseline methods for 3D tracking and motion forecasting, demonstrating how detailed map information improves accuracy. The datasets and methods are available at www.argoverse.org.Argoverse introduces two large-scale datasets for autonomous driving research: the Argoverse 3D Tracking dataset and the Argoverse Motion Forecasting dataset. These datasets are designed to support machine learning tasks such as 3D tracking and motion forecasting. The 3D Tracking dataset includes LiDAR measurements, 360° RGB video, front-facing stereo images, and 6-DOF localization, all aligned with detailed maps containing lane centerlines, drivable regions, and ground height. The Motion Forecasting dataset includes over 300,000 5-second tracked scenarios with identified vehicles for trajectory forecasting. Argoverse is the first dataset to include "HD maps" with 290 km of mapped lanes and geometric and semantic metadata. The data is released under a Creative Commons license, and the authors provide an API to connect map data with sensor information. The paper evaluates baseline methods for 3D tracking and motion forecasting, demonstrating how detailed map information improves accuracy. The datasets and methods are available at www.argoverse.org.