This repo includes four new real-world human activity recognition (HAR) datasets collected under federated learning settings, which first appear at the MobiSys 2021 paper: ClusterFL: A Similarity-Aware Federated Learning System for Human Activity Recognition .
The first dataset is a large-scale dataset collected using an Android App in a crowdsourcing manner. The other three are collected in indoor environments.
The four datasets are publicly available in the dropbox link . Each dataset is accompanied by a Python file "data_pre.py" for preprocessing and loading each node's data separately in federated learning. The data and processing file are compressed to a ".zip" file for each dataset. Please click the following links for more detail descriptions and downloading each dataset.
The datasets of this project are made available for non-commercial, academic research only. If you would like to use the datasets of this project, please cite the following papers:
@inproceedings{ouyang2021clusterfl,
title={ClusterFL: a similarity-aware federated learning system for human activity recognition},
author={Ouyang, Xiaomin and Xie, Zhiyuan and Zhou, Jiayu and Huang, Jianwei and Xing, Guoliang},
booktitle={Proceedings of the 19th Annual International Conference on Mobile Systems, Applications, and Services},
pages={54--66},
year={2021}
}
@article{ouyang2022clusterfl,
title={ClusterFL: A Clustering-based Federated Learning System for Human Activity Recognition},
author={Ouyang, Xiaomin and Xie, Zhiyuan and Zhou, Jiayu and Xing, Guoliang and Huang, Jianwei},
journal={ACM Transactions on Sensor Networks},
volume={19},
number={1},
pages={1--32},
year={2022},
publisher={ACM New York, NY}
}