An approach for local feature evaluation

In this paper, we introduce a unified feature evaluation toolbox, which contains several state of the art local features such as SIFT, SURF, KAZE, ORB, BRISK, and BIO. With our toolbox, we perform standard evaluation metrics (repeatability, precision-recall), but also the more practical assessment in terms of vision applications. In this version, we provide a refined evaluation methodology of image stitching and depth propagation. We will open the toolbox, and hope to make a various state of the art computer vision and robotics applications.