An active community for multi-sensor, multi-channel, annotated human activity datasets that are openly accessible and are continuously in development and discussion. This community attempts to bring forth accessible data structures and standards for data storage. The community attempts to focus on application-based datasets to bring in the complexity of transition windows, occlusion, and other such randomness within the dataset. This is done to tackle questions on transfer learning, annotation effort, annotation label development, automatic annotation techniques, transition window, and other such topics.  

To create a dataset for HAR, we recommend the guidelines of the following paper: "A Tutorial on Dataset Creation for Sensor-based Human Activity Recognition", PerCom, 2023, DOI: 10.1109/PerComWorkshops56833.2023.10150401

When publishing datasets, we recommend adhering to the FAIR principles: "The FAIR Guiding Principles for scientific data management and stewardship", 2016, DOI: 10.1038/sdata.2016.18