OoDIS
收藏资源简介:
OoDIS是由亚琛工业大学、伍珀塔尔大学和北京理工大学合作开发的数据集,专注于异常实例分割,旨在提高自动驾驶车辆对环境中未知物体的识别能力。该数据集整合了多个现有的异常分割数据集,如Fishyscapes、RoadAnomaly21和RoadObstacle21,提供了多样化的真实世界道路异常案例。创建过程中,研究团队对这些数据集进行了扩展和重新标注,以支持实例分割任务。OoDIS的应用领域主要集中在自动驾驶技术,特别是解决在训练中未见过的异常物体的精确分割问题,这对于避免交通事故至关重要。
OoDIS is a dataset developed collaboratively by RWTH Aachen University, the University of Wuppertal, and Beijing Institute of Technology. It focuses on anomalous instance segmentation, aiming to enhance the ability of autonomous vehicles to recognize unknown objects in their surrounding environments. This dataset integrates multiple existing anomaly segmentation datasets, such as Fishyscapes, RoadAnomaly21, and RoadObstacle21, providing diverse real-world road anomaly cases. During its development, the research team expanded and re-annotated these datasets to support the instance segmentation task. The main application scenarios of OoDIS are concentrated in autonomous driving technology, particularly addressing the accurate segmentation of anomalous objects unseen during model training, which is crucial for avoiding traffic accidents.




