randall-lab/kitti-stereo2012

KITTI Stereo 2012数据集是一个广泛用于评估自动驾驶场景中的立体视觉、光流和场景流算法的基准数据集。它提供了真实世界中的成对图像，这些图像是在城市和乡村环境中捕获的。数据集包括高分辨率的立体图像对、密集视差地面真实地图、校准文件，以及用于立体匹配研究的相关内容。该数据集使得立体算法的公平比较成为可能，并用作非遮挡和遮挡区域使用视差误差度量进行性能评估的基准。

The KITTI Stereo 2012 dataset is a widely used benchmark for evaluating stereo vision, optical flow, and scene flow algorithms in autonomous driving scenarios. It provides real-world image pairs captured in urban and rural environments, high-resolution stereo image pairs, dense disparity ground truth maps, and calibration files for geometric reconstruction. The dataset enables fair comparison of stereo algorithms and serves as a benchmark for performance evaluation on non-occluded and occluded regions using disparity error metrics.