WAYMO OPEN MOTION DATASET
收藏资源简介:
WAYMO OPEN MOTION DATASET是由Waymo LLC创建的大规模交互式运动预测数据集,旨在推动自动驾驶系统中的运动预测模型发展。该数据集包含超过100,000个场景,每个场景20秒长,覆盖1750公里的道路,数据来源于美国六个城市的车辆、行人和自行车之间的互动。数据集使用高精度的3D自动标注系统生成高质量的3D边界框,并提供相应的高清3D地图。此外,数据集还引入了新的度量标准,用于全面评估单个代理和联合代理交互运动预测模型的性能。该数据集的应用领域主要集中在提高自动驾驶系统中的运动预测准确性,特别是在需要预测多个对象联合行为的复杂场景中。
The Waymo Open Motion Dataset is a large-scale interactive motion prediction dataset created by Waymo LLC, aimed at advancing the development of motion prediction models in autonomous driving systems. It contains over 100,000 scenarios, each 20 seconds in duration, spanning 1,750 kilometers of roadways. The data is sourced from interactions between vehicles, pedestrians, and cyclists across six cities in the United States. The dataset is generated using a high-precision 3D automatic annotation system to produce high-quality 3D bounding boxes, and it also provides corresponding high-definition 3D maps. Additionally, the dataset introduces novel metrics for comprehensively evaluating the performance of individual agent and joint agent interactive motion prediction models. Its application scenarios mainly focus on improving the accuracy of motion prediction in autonomous driving systems, particularly in complex scenarios where the joint behavior of multiple objects needs to be predicted.

- 1Large Scale Interactive Motion Forecasting for Autonomous Driving : The Waymo Open Motion DatasetWaymo LLC · 2021年



