Campus & Shelf

该数据集名为CAMPUS，包含了四个场景的视频序列，每个场景由四台摄像机捕捉。这些视频序列中展现了丰富的人体姿态和活动，摄像机之间的视角有一定的重叠。此外，该数据集还提供了跟踪注释，以及针对动作和属性识别的新注释。它包含5个动作类别和9个属性类别的标注信息，适用于物体检测、多对象跟踪、动作识别和人体属性识别等多项任务。

This dataset, named CAMPUS, consists of video sequences across four scenarios, each captured by four cameras. The video sequences feature diverse human poses and activities, with partial overlapping fields of view among the cameras. In addition, this dataset provides tracking annotations as well as novel annotations for action and attribute recognition. It contains annotations for 5 action categories and 9 attribute categories, supporting multiple tasks including object detection, multi-object tracking, action recognition, and human attribute recognition.

搜集汇总

背景与挑战

背景概述

该数据集是用于多视角目标跟踪研究的综合数据集，包含CAMPUS、EPFL和PETS09三个子集，具有同步视频、全标注轨迹和跨视角一致ID分配的特点。数据集特别设计了密集前景、复杂场景和多样目标尺度等挑战性场景，并提供详细的标注规范和3D坐标转换方法。