San Francisco Landmark Dataset

该数据集包含了大约一百万张由汽车在旧金山市中心收集的全景图像中提取的透视图像，这些图像带有GPS测量的相机位置和方向信息。此外，数据集还包含了一个小型的测试数据集，其中包含了参考姿态，这允许评估微调的效果。整个数据集规模约为百万张图像，其任务是进行相对相机姿态估计。

This dataset contains approximately one million perspective images extracted from panoramic images collected by cars in downtown San Francisco, each with GPS-measured camera position and orientation information. Additionally, the dataset includes a small test dataset containing reference poses, which allows for evaluating the effect of fine-tuning. The total scale of the dataset is around one million images, and its task is relative camera pose estimation.