ISAR
收藏资源简介:
ISAR数据集是由苏黎世联邦理工学院的自主系统实验室创建,旨在支持单次和少量样本的对象实例分割与重识别研究。该数据集包含24个测试案例,总计84个训练和评估序列,每个序列长度为400帧,帧率为30Hz。数据集使用Habitat AI模拟器记录,场景来自Replica和Habitat-Matterport 3D研究数据集,对象来自YCB数据集。ISAR数据集特别设计用于测试和加速算法的发展,这些算法能够从单个或少数稀疏训练示例中鲁棒地检测、分割和重新识别对象,适用于机器人、增强现实和自动驾驶等空间AI应用。
The ISAR dataset was created by the Autonomous Systems Lab at ETH Zurich, aiming to support research on single-shot and few-shot object instance segmentation and re-identification. This dataset includes 24 test cases, totaling 84 training and evaluation sequences, each with a length of 400 frames and a frame rate of 30 Hz. Recorded using the Habitat AI simulator, its scenes are derived from the Replica and Habitat-Matterport 3D Research Datasets, while its objects are sourced from the YCB Dataset. The ISAR dataset is specifically designed to test and accelerate the development of algorithms that can robustly detect, segment, and re-identify objects from single or a small number of sparse training examples, and is suitable for spatial AI applications such as robotics, augmented reality, and autonomous driving.




