KITTI

该数据集是一个用于评估深度预测的标准基准，以激光雷达读数作为评估的地面真实值。数据划分采用了Eigen等人的方法，而预处理则是基于Zhou等人的工作。该数据集的规模包括39810个训练片段、4424个验证片段以及697个测试片段，其任务是对深度进行估计。

This dataset is a standard benchmark for depth prediction evaluation, using LiDAR readings as the ground truth for assessment. The data partitioning follows the method proposed by Eigen et al., and the preprocessing pipeline is based on the work of Zhou et al. The dataset comprises 39,810 training clips, 4,424 validation clips, and 697 test clips, with its primary task being depth estimation.