Homography with Degree

We selected 100 images from MF-DFC22 dataset and divided them into four categories: architecture, mountains, coasts, and farmland. We performed different degrees of projection transformations in eight directions on each image to explore the performance of different keypoint detector under projection transformations.

我们从MF-DFC22数据集中选取100张图像，并将其划分为建筑、山地、海岸与农田四个类别。我们对每张图像在八个方向上施加不同程度的投影变换，以探究不同关键点检测器（keypoint detector）在投影变换条件下的性能表现。