KITTI-Pseudo-Depth
收藏资源简介:
该数据集为KITTI原始数据集提供了高质量的伪深度图,这些深度图是使用先进的Depth Anything V2模型生成的。它旨在促进单目深度估计、3D目标检测以及自动驾驶应用领域的研究。数据集遵循标准的Eigen划分方式,包含训练集、验证集和测试集。原始图像来源于KITTI视觉基准测试套件(原始数据),深度图由Depth Anything V2模型生成。数据格式方面,RGB图像为.png格式,深度图为.npy格式(float32精度,以实现零损失精度)。
This dataset provides high-quality pseudo-depth maps for the KITTI raw dataset, generated using the advanced Depth Anything V2 model. It aims to facilitate research in monocular depth estimation, 3D object detection, and autonomous driving applications. The dataset follows the standard Eigen split, including training, validation, and test sets. Original images are sourced from the KITTI vision benchmark suite (raw data), and depth maps are generated by the Depth Anything V2 model. Regarding data formats, RGB images are in .png format, and depth maps are in .npy format (float32 precision for lossless accuracy).
数据集概述:KITTI Pseudo Depth (Eigen Split) with Depth Anything V2
这是一个为 KITTI 原始数据集提供高质量伪深度图的数据集,使用 Depth Anything V2 模型生成,旨在支持单目深度估计、3D 目标检测和自动驾驶相关研究。
任务类别
- 深度估计
- 图像到图像
语言
- 英语
标签
- kitti, depth-anything-v2, monocular-depth-estimation, autonomous-driving
数据集规模
- 样本数量: 10,000 < N < 100,000
数据划分
- 采用标准的 Eigen Split 划分(训练、验证、测试)。
数据格式
- 原始图像: RGB 图片,格式为
.png - 深度图: 深度图文件格式为
.npy(float32 类型),保证零精度损失。
数据来源
- 原始图像: KITTI Vision Benchmark Suite(原始数据)
- 深度生成器: Depth Anything V2
引用与致谢
使用本数据集时,请引用以下文献:
-
Depth Anything V2:
@article{depth_anything_v2, title={Depth Anything V2}, author={Yang, Lihe and Kang, Bingyi and Huang, Zilong and Zhao, Zhen and Xu, Xiaogang and Feng, Jiashi and Zhao, Hengshuang}, journal={arXiv preprint arXiv:2406.09414}, year={2024} }
-
KITTI 数据集:
@article{Geiger2013IJRR, author = {Andreas Geiger and Philip Lenz and Christoph Stiller and Raquel Urtasun}, title = {Vision meets Robotics: The KITTI Dataset}, journal = {International Journal of Robotics Research (IJRR)}, year = {2013} }
@inproceedings{Menze2015CVPR, author = {Moritz Menze and Andreas Geiger}, title = {Object Scene Flow for Autonomous Vehicles}, booktitle = {Conference on Computer Vision and Pattern Recognition (CVPR)}, year = {2015} }




