Weakly Supervised Person Dataset (WSPD)
收藏arXiv2020-05-01 更新2024-06-21 收录
下载链接：
https://github.com/cvpaperchallenge/FashionCultureDataBase_DLoader
下载链接
链接失效反馈官方服务：
资源简介：
Weakly Supervised Person Dataset (WSPD)是由日本国立先进工业科学技术研究所创建的大规模人物检测数据集，包含870多万张图像及其边界框。该数据集通过半自动图像收集和数据精炼过程构建，利用社交媒体服务（如Flickr）的图像进行人物检测和误报抑制。WSPD数据集旨在通过其大规模和高多样性的特点，提升人物检测算法的性能，特别是在自动驾驶、移动机器人和无人机等领域的应用。
The Weakly Supervised Person Dataset (WSPD) is a large-scale person detection dataset created by the National Institute of Advanced Industrial Science and Technology (AIST) of Japan, which encompasses over 8.7 million images and their associated bounding boxes. This dataset is constructed via a semi-automated image collection and data refinement process, leveraging images from social media services such as Flickr for person detection and false positive suppression. The WSPD dataset aims to enhance the performance of person detection algorithms through its large scale and high diversity, particularly for applications in autonomous driving, mobile robotics, unmanned aerial vehicles (UAVs), and other relevant fields.
提供机构：
日本国立先进工业科学技术研究所
创建时间：
2020-03-27
