WiFi Signals Can Now Track You Through Walls
A new open-source project called WiFi-DensePose can track human movement and poses through walls using only the signals from a standard WiFi router. The project, which has over 32,000 stars on GitHub, uses signal processing to create a 3D representation of a person's body without any cameras, opening up new possibilities for privacy-centric monitoring and interaction.
The open-source project builds upon foundational research from Carnegie Mellon University, which first demonstrated that a deep neural network could map the phase and amplitude of WiFi signals to the UV coordinates of 24 different human body regions. This academic work proved that with enough training data, WiFi's Channel State Information (CSI) could achieve human pose estimation with performance comparable to image-based methods. The key technology, Channel State Information (CSI), is metadata that describes how a WiFi signal travels from a transmitter to a receiver. Human bodies reflect, absorb, and scatter these radio waves, causing unique disturbances in the signal's amplitude and phase. A neural network is trained to recognize the specific CSI patterns that correspond to different human poses, effectively learning to "see" without a camera. This implementation, created by a developer known as "ruvnet," aims to be a production-ready system, distinguishing it from purely academic research. It's designed to run on the edge, using inexpensive hardware like ESP32-S3 microcontroller boards, and offers both a Python implementation for rapid prototyping and a Rust version for an 800x performance speedup in production. For developers, the project is designed to be accessible. The Python version includes a FastAPI backend with REST and WebSocket APIs for querying pose data and streaming it in real-time. The system can output not just pose coordinates, but also presence detection and even vital signs like breathing and heart rate by analyzing the micro-Doppler shifts in the WiFi signals. While often touted as "privacy-preserving" due to the lack of cameras, the technology introduces a new set of ethical considerations. The ability to monitor movement, sleep patterns, and daily routines without consent turns wireless infrastructure into a potential surveillance layer. This has led to discussions about "inference fuel," where even anonymized motion data could be used to build behavioral profiles for commercial or monitoring purposes. The project's GitHub repository provides documentation for setting up the system using either specialized hardware like Intel 5300 or Atheros network cards, or more accessible ESP32-S3 boards. The output is a stream of pose data, such as UV coordinates for body parts, which can then be used as an input for other applications, from patient monitoring in healthcare to character animation. However, there has been some debate within the developer community regarding the project's maturity. Some have raised concerns about the completeness of the code and the validity of its claimed capabilities, suggesting it may be more of a proof-of-concept than a fully production-ready system. Ultimately, WiFi-based sensing represents a significant shift in how developers can approach human-computer interaction and environmental awareness. It opens the door for applications in spaces where cameras are intrusive or impractical, such as in-home elderly care or post-disaster search and rescue. The technology's ability to "see" through solid objects like rubble makes it a potentially life-saving tool in emergency scenarios.
