UW IOM (University of Washington Indoor Object Manipulation)

华盛顿大学室内物体操纵 (UW IOM) 数据集包含 20 名 18-25 岁年龄组参与者的视频（和相应的骨骼跟踪信息），其中 15 名是男性，其余 5 名是女性。这些视频是使用用于 Xbox One 的 Kinect 传感器以平均每秒 12 帧的速率录制的。每个参与者执行相同的任务，从三个不同的垂直架子上拾起六个物体（三个相同的空盒子和三个相同的杆），将它们放在桌子上，然后将它们放回从它们被拾起的架子上，然后从中间的架子上提着盒子走出现场。盒子用双手操作，而杆只用一只手操作。上述任务以相同的顺序重复 3 次，使得每个视频的持续时间约为 3 分钟。我们将动作分类为 17 个标签，其中每个标签遵循四层层次结构。第一层表示是否操纵了盒子或杆，第二层表示人体运动（步行、站立和弯曲），第三层捕获对象操作的类型（如果适用）（伸手、拿起、放置和握住） )，第四层表示进行操作的表面的相对高度（低、中和高）。

The University of Washington Indoor Object Manipulation (UW IOM) Dataset contains videos and corresponding skeletal tracking data from 20 participants aged 18 to 25 years old, including 15 males and 5 females. These videos were recorded using a Kinect sensor for Xbox One at an average frame rate of 12 frames per second. Each participant performed the identical task: picking up six objects (three identical empty boxes and three identical rods) from three separate vertical shelves, placing them onto a table, returning them to the shelves they were collected from, and finally carrying the boxes out of the scene via the middle shelf. Boxes are manipulated with both hands, while rods are manipulated with only one hand. The aforementioned task was repeated three times in the same sequence, resulting in each video having an approximate duration of 3 minutes. We categorized the actions into 17 labels, each following a four-level hierarchical structure. The first level indicates whether a box or rod was manipulated; the second level represents human body movements such as walking, standing, and bending; the third level captures the type of object manipulation (if applicable) including reaching, picking up, placing, and holding; and the fourth level denotes the relative height of the surface where the operation was performed, with three tiers: low, medium, and high.

搜集汇总

背景与挑战

背景概述

UW IOM数据集包含20名18-25岁参与者的视频和骨骼跟踪数据，使用Kinect传感器以12fps录制，记录他们执行重复的室内物体操纵任务，如拾起和放置盒子和杆。动作被分类为17个标签，基于四层层次结构，涵盖物体类型、人体运动、操作类型和表面高度。