WiSe4Car: A Dataset for WiFi Sensing in Automotive Vehicles

WiFi sensing, i.e., using WiFi signals to scan the environment, is a promising technology with a plurality of potential applications in the automotive domain. In addition to being cost-effective, since WiFi sensing does not rely on images, it addresses privacy concerns and/or regulations in vehicle interiors. However, there is a notorious lack of publicly available datasets necessary to train the underlying machine learning (ML) models and further develop the technology in this domain. In this paper, we propose one such dataset consisting of CSI (Channel State Information) recordings in more than 100 different scenarios involving multiple individuals and different vehicles. We have not only labeled CSI data by participants’ height, weight, gender, and seat position among others, but also include anonymized video recordings for further labeling in the future (say, for example, activity, etc.). In addition, CSI data was collected every time from multiple spots within the vehicle allowing for later studies concerning sensor positioning in an application of interest. The proposed dataset is made publicly available and intended to help establishing WiFi sensing to be used within automotive vehicles, which we demonstrate by an exemplary case study.