SaRNet

SaRNet是由纽约大学电气与计算机工程系创建的一个针对深度学习辅助搜索和救援的卫星影像数据集。该数据集包含2552张高分辨率卫星图像，共有4206个轴对齐的边界框，用于标记潜在的目标对象。数据集的创建过程涉及500多名志愿者在实时搜索环境中标记潜在目标，特别是在搜索一名失踪的滑翔翼飞行员时收集的数据。SaRNet的应用领域主要集中在搜索和救援任务，特别是在人道主义救援中，旨在通过深度学习技术提高搜索效率和准确性。

SaRNet is a satellite imagery dataset for deep learning-aided search and rescue, developed by the Department of Electrical and Computer Engineering at New York University. This dataset contains 2,552 high-resolution satellite images, with a total of 4,206 axis-aligned bounding boxes used to annotate potential target objects. The creation of this dataset involved over 500 volunteers annotating potential targets in real-time search scenarios, with data collected specifically during the search for a missing hang glider pilot. The primary application domain of SaRNet centers on search and rescue missions, especially humanitarian relief efforts, with the goal of enhancing search efficiency and accuracy through deep learning technologies.