Transfer learning for CSI refers to the application of machine learning models trained on channel state information data from one environment, device configuration, or deployment scenario to a different but related target domain, reducing or eliminating the need to collect large volumes of labeled training data in every new setting. This approach is particularly important in CSI-based sensing research because WiFi signals are highly sensitive to environmental geometry, hardware characteristics, and user placement, meaning models trained in one context frequently fail to generalize to another — a core challenge highlighted in studies of environment-dependent crowd counting and occupancy detection. Key variants include domain adaptation, where a model is fine-tuned using limited target-domain samples, and domain-invariant feature learning, where representations are engineered or learned to be robust across environments, devices, or building layouts.
Source Papers
- CrossSense: Towards Cross-Site and Large-Scale WiFi Sensing ↗ — CrossSense: Towards Cross-Site and Large-Scale WiFi Sensing
- Device-free occupancy detection and crowd counting in smart buildings with WiFi-enabled IoT ↗ — Device-free occupancy detection and crowd counting in smart
- Investigation of Environment Dependence in Wi-Fi CSI-Based Crowd Counting Systems ↗ — Investigation of Environment Dependence in Wi-Fi CSI-Based C
- SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing ↗ — SenseFi: A library and benchmark on deep-learning-empowered
- Towards Environment Independent Device Free Human Activity Recognition ↗ — Towards Environment Independent Device Free Human Activity R
- WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired by Dynamic Topology Structure ↗ — WiGNN: WiFi-Based Cross-Domain Gesture Recognition Inspired