Occluded-DukeMTMC

该数据集是针对遮挡人物重识别任务最具挑战性的数据集之一，原因在于其包含了多样化的场景和干扰因素。它旨在评估深度学习模型在遮挡场景下的性能。数据集的规模包括：702个人的15,618张训练图像；519个人的2,210张查询图像；以及1,110个人的17,661张图库图像。This dataset is one of the most challenging datasets for the task of occluded person re-identification, as it encompasses diverse scenarios and interfering factors. It aims to evaluate the performance of deep learning models under occluded scenarios. The scale of the dataset is as follows: 15,618 training images from 702 individuals; 2,210 query images from 519 individuals; and 17,661 gallery images from 1,110 individuals.

搜集汇总

背景与挑战

背景概述

Occluded-DukeMTMC是一个用于遮挡行人重识别问题的数据集，基于DukeMTMC-reID数据集重新划分生成。其关键特点是所有查询图像和10%的图库图像均为遮挡行人图像，确保在计算图像对距离时至少包含一张遮挡图像，专门用于研究遮挡场景下的行人重识别任务。