Ithaca365
收藏资源简介:
Ithaca365数据集由康奈尔大学创建,专注于在重复和挑战性天气条件下自动驾驶感知的研究。该数据集通过在15公里的路线上重复记录数据,涵盖了城市、高速公路、乡村和校园等多种场景,以及雪、雨、阳光等多种天气条件。数据集包含超过68万帧的LiDAR、图像和GPS数据,并提供了道路和对象的注释,包括3D边界框和模态掩码,以捕捉部分遮挡。Ithaca365数据集的独特之处在于其重复路线的设计,这为对象发现、持续学习和异常检测等研究方向开辟了新的可能性。此外,数据集还引入了新的评估指标,用于正确评估道路和背景类别的模态分割性能,展示了其在自动驾驶感知任务中的实用性和重要性。
The Ithaca365 dataset, developed by Cornell University, focuses on autonomous driving perception research under repetitive and challenging weather conditions. It collects data by repeatedly traversing a 15-kilometer route, covering diverse scenarios including urban areas, highways, rural roads, and campuses, as well as various weather conditions such as snow, rain, and sunlight. The dataset contains over 680,000 frames of LiDAR, image, and GPS data, and provides annotations for roads and objects, including 3D bounding boxes and modal masks to capture partial occlusions. What distinguishes Ithaca365 is its repeated route design, which unlocks new research opportunities for tasks such as object discovery, continual learning, and anomaly detection. Additionally, the dataset introduces novel evaluation metrics to accurately assess the modal segmentation performance of road and background categories, demonstrating its practicality and significance in autonomous driving perception tasks.




