A Vision Transformer (ViT) is a deep learning architecture that applies the transformer mechanism, originally developed for natural language processing, directly to image or image-like data by dividing inputs into fixed-size patches and processing them through self-attention layers to capture long-range spatial dependencies. In the context of WiFi CSI sensing, ViTs are used to model complex spatiotemporal patterns in CSI feature representations — such as heatmaps or spectrograms of channel amplitude and phase — enabling accurate recognition of activities, gestures, and human presence. Key variants relevant to this field include hybrid CNN-Transformer models that combine local feature extraction with global attention, as well as lightweight or distilled ViT variants that are of particular interest for green sensing applications where computational efficiency and reduced energy consumption are critical constraints.
Source Papers
- A Survey on Green Wireless Sensing: Energy-Efficient Sensing via WiFi CSI and Lightweight Learning ↗ — A Survey on Green Wireless Sensing: Energy-Efficient Sensing
- 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
- SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing ↗ — SenseFi: A library and benchmark on deep-learning-empowered