人车物目标关联分析模型实验数据集

本数据集主要用于研究验证人车物目标关联分析模型所涉及的基于图卷积的多摄像头多目标跟踪技术和基于稀疏图卷积模型的多类型目标关联性分析及轨迹预测技术相关性能指标，包括多目标追踪准确率MCTA、目标跟踪准确度分数MOTA、多目标跟踪精确度分数MOTP、目标检测准确率Precision、最小平均位移误差mADE、最小预测值mFDE、目标追踪准确率IDF1等。 数据集包括5个文件夹：MCT数据文件、CityFlow数据文件、SDD数据文件、ETH数据文件、UCY数据文件，存放了用于测试验证人车物目标关联分析模型所涉及的多摄像头多目标跟踪技术和多类型目标关联性分析及轨迹预测技术所需要的实验数据。 数据量共98.84 GB，其中MCT数据集：0.99GB；CityFlow数据集：31.5GB；ETH数据集：81.72GB；UCY数据集：0.15GB；Stanford Drone Dataset (SDD)数据集：66.1 GB。This dataset is primarily designed for studying and validating the performance metrics of graph convolution-based multi-camera multi-object tracking technology and sparse graph convolution-based multi-type target association analysis and trajectory prediction technologies employed in human, vehicle, and object target association analysis models. The included performance metrics are: Multi-Camera Tracking Accuracy (MCTA), Multiple Object Tracking Accuracy (MOTA), Multiple Object Tracking Precision (MOTP), Target Detection Precision, minimum Average Displacement Error (mADE), minimum Final Displacement Error (mFDE), and Identity F1 Score (IDF1), among others. The dataset comprises 5 folders, namely MCT data files, CityFlow data files, SDD data files, ETH data files, and UCY data files, which store the experimental data required for testing and validating the aforementioned multi-camera multi-object tracking technology, multi-type target association analysis, and trajectory prediction technologies for human, vehicle, and object target association analysis models. The total storage size of the entire dataset is 98.84 GB, with a breakdown as follows: MCT dataset: 0.99 GB; CityFlow dataset: 31.5 GB; ETH dataset: 81.72 GB; UCY dataset: 0.15 GB; Stanford Drone Dataset (SDD): 66.1 GB.

搜集汇总

背景与挑战

背景概述

该数据集专为验证人车物目标关联分析模型而设计，重点支持基于图卷积的多摄像头多目标跟踪技术以及多类型目标关联分析与轨迹预测技术的研究。它包含了来自MCT、CityFlow、SDD、ETH和UCY五个子集的实验数据，总容量为98.84 GB。