SEVD
收藏资源简介:
SEVD是一个创新的合成事件基础视觉数据集,由亚利桑那州立大学创建,专为自主驾驶和交通监控任务设计。该数据集包含27小时的固定感知数据和31小时的自我感知数据,总计超过900万个边界框,覆盖多种环境和不同参数。SEVD利用CARLA模拟器记录数据,涵盖城市、郊区、乡村和高速公路场景,以及多种光照和天气条件。此外,数据集还包括RGB图像、深度图、光流、语义和实例分割,以及GNSS和IMU数据,为场景理解提供全面支持。SEVD的应用领域包括交通参与者检测和自主驾驶系统的评估,旨在解决动态视觉数据处理和实时响应的挑战。
SEVD is an innovative synthetic event-based visual dataset developed by Arizona State University, specifically tailored for autonomous driving and traffic monitoring tasks. This dataset includes 27 hours of stationary perception data and 31 hours of ego-centric perception data, with over 9 million bounding boxes in total, covering diverse environments and varying parameters. SEVD collects data using the CARLA simulator, encompassing urban, suburban, rural, and highway scenarios, as well as diverse lighting and weather conditions. Furthermore, the dataset provides RGB images, depth maps, optical flow, semantic and instance segmentation data, along with GNSS and IMU data, offering comprehensive support for scene understanding. The application domains of SEVD cover traffic participant detection and autonomous driving system evaluation, with the goal of addressing the challenges in dynamic visual data processing and real-time response.




