RUBIK
收藏资源简介:
RUBIK数据集是由法国桥梁公路力学学院和法国波尔多大学的研究人员创建的,基于nuScenes数据集的图像。这个数据集特别设计用于细粒度评估图像匹配方法在几何挑战方面的局限性。它包含16500对图像,分布在33个不同的难度等级中,是根据场景重叠、尺度比和视角差异三个标准来组织的。数据集的创建是为了提供一个更全面的评估,以推动相机位姿估计方法的改进。
The RUBIK dataset was created by researchers from the French National School of Bridges, Roads and Civil Engineering Mechanics and the University of Bordeaux, based on images from the nuScenes dataset. This dataset is specifically tailored for fine-grained assessment of the limitations of image matching methods in the context of geometric challenges. It comprises 16,500 image pairs, which are categorized into 33 distinct difficulty levels based on three criteria: scene overlap, scale ratio, and viewpoint disparity. The dataset was developed to enable a more comprehensive evaluation, thereby promoting the improvement of camera pose estimation methods.

- 1RUBIK: A Structured Benchmark for Image Matching across Geometric Challenges法国桥梁公路力学学院, 法国波尔多大学 · 2025年



