DoPose-6D
收藏资源简介:
DoPose-6D数据集由德国多特蒙德工业大学创建,专注于6D姿态估计和对象分割,特别适用于智能机器人抓取和操作场景。该数据集包含两种场景:箱内拾取和桌面,主要用于箱内拾取场景。数据集包含18种不同形状、大小、颜色和纹理的对象,通过高动态范围图像捕捉,提供丰富的真实图像和标注,适用于多种计算机视觉任务。创建过程中,使用了Zivid Two结构光相机和Kuka iiwa LBR 14机器人手臂进行数据采集和标注,确保数据的高质量和准确性。该数据集的应用领域包括机器人抓取、自动化仓库系统等,旨在解决高杂乱度和高遮挡环境下的对象识别和抓取问题。
DoPose-6D Dataset is developed by the Technical University of Dortmund, Germany, focusing on 6D pose estimation and object segmentation, and is specifically applicable to intelligent robotic grasping and manipulation scenarios. It includes two scenarios: bin picking and tabletop, with the bin picking scenario being the main focus. The dataset contains 18 objects with distinct shapes, sizes, colors and textures, which are captured using high dynamic range (HDR) images, providing rich real-world images and corresponding annotations suitable for various computer vision tasks. During the dataset creation process, a Zivid Two structured-light camera and a Kuka iiwa LBR 14 robotic arm were employed for data acquisition and annotation, ensuring high data quality and accuracy. The application fields of this dataset include robotic grasping, automated warehouse systems and others, aiming to solve the problems of object recognition and grasping in highly cluttered and heavily occluded environments.




