KITTI Object Detection Dataset

The authors present a KITTI Object Detection dataset (is a part of a larger KITTI dataset) obtained from a VW station wagon for application in mobile robotics and autonomous driving research. They meticulously recorded 6 hours of traffic scenarios at 10–100 Hz, utilizing a range of sensor modalities, including high-resolution color and grayscale stereo cameras, a Velodyne 3D laser scanner, and a high-precision GPS/IMU inertial navigation system. The dataset encompasses diverse scenarios, capturing real-world traffic situations across freeways, rural areas, and inner-city scenes with numerous static and dynamic objects. The data is calibrated, synchronized, timestamped, and provided in rectified and raw image sequences. Additionally, the dataset contains object labels in the form of 3D tracklets, and the authors offer online benchmarks for stereo, optical flow, object detection, and other tasks. The paper details the recording platform, data format, and utilities provided by the authors.

作者们呈现了一套来自VW旅行车的KITTI目标检测数据集（该数据集为更大规模KITTI数据集的一部分），旨在应用于移动机器人和自动驾驶研究。他们严谨地记录了6小时的交通场景，采样频率为10至100赫兹，并采用了多种传感器模式，包括高分辨率彩色和灰度立体相机、Velodyne 3D激光扫描仪以及高精度GPS/IMU惯性导航系统。该数据集涵盖了多样化的场景，捕捉了跨越高速公路、乡村地区及市区场景中的真实交通状况，其中包含了众多静态和动态对象。数据经过校准、同步、时间戳标记，并以校正和原始图像序列的形式提供。此外，数据集包含以3D轨迹段形式的对象标签，且作者们提供了在线基准测试，涉及立体视觉、光流、目标检测以及其他任务。论文详细描述了记录平台、数据格式及作者们提供的工具。