MSMT17

该数据集名为MSMT17，是一个大规模的人体重识别数据集，包含了15个摄像头捕捉的4,101个身份的126,441个边界框。其中，32,621张图像，涉及1,041个身份用于训练。该数据集的规模达到了4,101个身份和126,441个边界框，任务类型为无监督人体重识别。

The dataset named MSMT17 is a large-scale person re-identification dataset. It contains 126,441 bounding boxes belonging to 4,101 unique identities, captured by 15 cameras. Of these, 32,621 images involving 1,041 identities are used for training. With a total scale of 4,101 identities and 126,441 bounding boxes, the task type of this dataset is unsupervised person re-identification.

搜集汇总

背景与挑战

背景概述

MSMT17是一个大规模无监督人体重识别数据集，包含15个摄像头拍摄的4,101个身份的126,441个边界框，其中32,621张图像（1,041个身份）用于训练。该数据集以其多样本和跨摄像头特性为人体重识别研究提供了丰富资源。