MVB (Multi View Baggage)

MVB (多视图行李) 是行李里德任务的数据集，与人里德有一些本质的区别。MVB的功能有三个方面。首先，MVB是第一个公开发布的大规模数据集，其中包含4519行李身份和22660带注释的行李图像以及其表面材料标签。其次，所有行李图像均由专门设计的多视图摄像头系统捕获，以处理姿势变化和遮挡，以获得尽可能完整的行李表面3D信息。第三，考虑到行李可能具有非常相似的外观，而数据是在两个真实的机场环境中收集的，成像因素彼此之间差异很大，因此MVB具有显着的舱内相似性和舱内不相似性。

MVB (Multi-View Baggage) is a dataset for baggage re-identification, which has several fundamental differences from person re-identification. MVB has three core characteristics. First, it is the first publicly released large-scale dataset that includes 4519 unique baggage identities and 22,660 annotated baggage images paired with their surface material labels. Second, all baggage images are captured by a specially designed multi-view camera system to address pose variations and occlusions, so as to obtain as complete 3D information of the baggage surface as possible. Third, considering that baggage may have highly similar appearances, and the data was collected in two real airport environments with largely different imaging factors, MVB exhibits significant intra-class similarity and inter-class dissimilarity.

搜集汇总

背景与挑战

背景概述

MVB是一个专注于行李里德任务的大规模数据集，包含4519个行李身份和22660张带注释的图像，并带有表面材料标签，是首个公开发布的此类数据集。数据通过多视图摄像头系统捕获，以处理姿势变化和遮挡，获取完整3D信息，并在两个真实机场环境中收集，具有显著的舱内相似性和不相似性，适用于行李识别研究。