NIPD

NIPD数据集是由电子科技大学信息与软件工程学院创建的，专门用于非独立同分布（non-IID）的人体检测研究。该数据集包含从五个不同摄像头收集的10000张图像，每张图像至少包含一个人，总共有55,160个人体标注。数据集的创建过程包括自动过滤和标注，随后进行人工校验和精确标注。NIPD数据集主要用于解决IoT设备在实际应用中遇到的非IID数据分布问题，特别是在智能城市的监控领域中，通过联邦学习（FL）提升模型在不同设备和场景下的适应性和准确性。

The NIPD dataset was developed by the School of Information and Software Engineering, University of Electronic Science and Technology of China, and is purpose-built for research on non-independent and identically distributed (non-IID) human detection. It comprises 10,000 images collected from five distinct cameras, each containing at least one person, with a total of 55,160 human body annotations. The dataset creation workflow includes automatic filtering and initial annotation, followed by manual verification and precise final annotation. Primarily, the NIPD dataset is designed to address non-IID data distribution challenges faced by IoT devices in real-world applications, particularly in the smart city surveillance domain, to enhance the adaptability and accuracy of models across various devices and scenarios through Federated Learning (FL).