Indoor Pedestrian Tracking dataset

该数据集包含了从两个室内传感器节点收集的数据，涵盖了3248帧画面，其中最多有九名行人和一个病床，总计22557个标记对象。数据集特色在于三种场景：高行人流量、详细的跟踪分析以及病床与行人之间的互动。评估指标包括精确度、召回率和平均距离误差。该数据集的规模为3248帧，平均每帧有7.04个对象。其任务是对室内环境中的对象进行检测和跟踪。

This dataset contains data collected from two indoor sensor nodes, encompassing 3248 video frames. Within these frames, there are up to nine pedestrians and one hospital bed, with a total of 22557 annotated objects. This dataset features three distinct scenarios: high pedestrian traffic, detailed tracking analysis, and interactions between hospital beds and pedestrians. Evaluation metrics include precision, recall, and average distance error. The dataset has an average of 7.04 objects per frame. The core task of this dataset is object detection and tracking in indoor environments.

搜集汇总

背景与挑战

背景概述

该数据集是'MVSLab-IndoorCooperativePerception'项目的一部分，专注于室内行人跟踪，用于增强室内移动性。数据集以ROS bag格式存储，包含相机参数和传感器变换，支持Rviz可视化，标注采用CVAT工具，以BEV（鸟瞰图）格式的定向边界框进行标注，适用于实时、延迟感知的协同感知研究。