UT-HAR is a publicly available WiFi CSI dataset designed for human activity recognition (HAR), collected using commercial WiFi hardware and containing labeled CSI samples across multiple activity categories. It serves as a widely adopted benchmark in the field, enabling standardized evaluation and comparison of deep learning models — including CNNs, RNNs, and transformer-based architectures — for CSI-based sensing tasks, as demonstrated in benchmarking efforts such as SenseFi. Its public availability and structured format make it a key resource for reproducibility studies, helping researchers validate and replicate sensing pipelines without requiring custom data collection infrastructure.
Source Papers
- A survey on CSI-based Wi-Fi sensing datasets and models with a focus on reproducibility ↗ — A survey on CSI-based Wi-Fi sensing datasets and models with
- Deep Learning-Enhanced Human Sensing with Channel State Information: A Survey ↗ — Deep Learning-Enhanced Human Sensing with Channel State Info
- SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing ↗ — SenseFi: A library and benchmark on deep-learning-empowered