DAVE
收藏资源简介:
DAVE数据集由马里兰大学等机构创建,旨在为复杂和不可预测环境中的感知方法提供评估基准,特别是针对易受伤害的道路使用者(VRUs)。该数据集包含1231个视频片段,涵盖了16种不同的参与者类别和16种动作类型,标注了超过1300万个边界框,其中160万个框同时标注了参与者身份和动作/行为细节。数据集视频基于多种因素收集,如天气条件、时间、道路场景和交通密度。DAVE数据集的应用领域包括自动驾驶、视频跟踪、检测、时空动作定位等,旨在解决现有数据集在复杂和不可预测环境中的不足,提升感知算法的鲁棒性和准确性。
The DAVE dataset was developed by institutions including the University of Maryland, aiming to provide an evaluation benchmark for perception methods in complex and unpredictable environments, especially for Vulnerable Road Users (VRUs). This dataset contains 1231 video clips, covering 16 distinct participant categories and 16 action types, with over 13 million bounding boxes annotated. Among these, 1.6 million boxes are simultaneously annotated with participant identity and action/behavior details. The videos in the dataset are collected based on multiple factors including weather conditions, time of day, road scenarios and traffic density. Application fields of the DAVE dataset include autonomous driving, video tracking, object detection, spatiotemporal action localization and more. It is designed to address the limitations of existing datasets in complex and unpredictable environments, and improve the robustness and accuracy of perception algorithms.

- 1DAVE: Diverse Atomic Visual Elements Dataset with High Representation of Vulnerable Road Users in Complex and Unpredictable Environments马里兰大学、弗吉尼亚大学、波士顿大学 · 2024年



