2026 · Developer / Researcher
RuView
WiFi DensePose Human Pose Estimation
- Discipline
- ResearchMachine learningCoding
- For
- Study
- Stack
- WiFi SensingDensePoseComputer VisionSignal ProcessingPythonDeep Learning
Problem
Cameras are invasive. Can ordinary WiFi see human pose instead?
What I built
A signal-processing pipeline that turns commodity WiFi channel data into pose and presence estimates.
Result
Camera-free pose and presence detection on standard hardware.
Researching commodity WiFi-based human pose estimation — no camera required
What went into it
- Explored cutting-edge research in WiFi-based human pose estimation, leveraging commodity WiFi signals to detect human presence and body pose without cameras.
- Researched DensePose techniques applied to WiFi signal patterns for real-time human pose estimation, vital sign monitoring, and presence detection.
- Developed a privacy-preserving alternative to camera-based pose estimation, addressing surveillance concerns while maintaining detection accuracy.
- Implemented signal processing pipelines to extract meaningful human movement data from standard WiFi hardware, bridging networking and computer vision.
How I thought about it
Privacy was the constraint, not an afterthought. Most of the work was signal cleaning; the model is only as good as the noise you remove.