行人跟踪序列视频数据集

行人跟踪序列视频数据集是用于视频行人重识别任务的大规模数据集，原始数据包括12个户外摄像头，3个室内摄像头的180小时视频序列。该数据集包括训练集、验证集和测试集三部分。其中训练集包括842个行人的550419张视频帧，验证集包括200个行人的155191张视频帧，测试集包括2730个行人的2277075张视频帧。原始视频数据通过手机摄像头和部分相机采集，行人检测跟踪利用现有算法完成。行人ID标注工作由10个专业标注人员耗时2个月完成。

The pedestrian tracking sequence video dataset is a large-scale dataset designed for the video-based person re-identification task. The raw data consists of 180 hours of video sequences captured by 12 outdoor cameras and 3 indoor cameras. This dataset is divided into three subsets: training set, validation set, and test set. The training set contains 550,419 video frames from 842 distinct pedestrians, the validation set contains 155,191 video frames from 200 distinct pedestrians, and the test set contains 2,277,075 video frames from 2,730 distinct pedestrians. The raw video data was collected using smartphone cameras and several cameras, and pedestrian detection and tracking were implemented using existing algorithms. The pedestrian ID annotation work was completed by 10 professional annotators over a period of two months.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于视频行人重识别的大规模数据集，包含12个户外和3个室内摄像头的180小时视频序列，分为训练集、验证集和测试集，总计涵盖3772个行人的约298万张视频帧。其特点是数据量庞大（10.32GB）、场景覆盖广泛（室内外混合），并经过专业人工标注，适用于计算机视觉领域的目标重识别任务研究。