Development · GitHub repo
RuView
Real-time human pose, presence, and vital sign sensing using commodity WiFi.
RuView is an open-source spatial intelligence system that transforms ordinary WiFi radio signals into presence detection, vital sign tracking, and body pose estimation without using cameras. It reads Channel State Information (CSI) disturbances from low-cost ESP32 microcontrollers or research NICs to detect motion, breathing, heart rate, and human activity through walls.
What it's for
- Monitor breathing and heart rates without cameras or wearable devices
- Track room occupancy and presence through walls for smart home automation
- Detect falls and unusual immobility for non-invasive elderly care
- Estimate basic body pose skeletons using raw WiFi Channel State Information
Features
- Contactless vital sign monitoring measuring breathing (6–30 BPM) and heart rate (40–120 BPM)
- Camera-free 17-keypoint human pose estimation from WiFi CSI disturbances
- Through-wall presence detection and occupancy counting for up to 10 people
- Fall detection using phase-acceleration analysis with under 200 ms latency
- Smart home integrations supporting Home Assistant (MQTT), Apple HomeKit, Google Home, and Matter
- Edge-ready firmware running on inexpensive ESP32-S3 or ESP32-C6 boards
- Lightweight quantized machine learning models fitting in 8 KB for microcontrollers and single-board computers
- Local-first architecture requiring no cloud connectivity or external subscriptions
Pricing
Open source. Free and open source (MIT license); optional hardware sensors cost $6–$140
Details
- Type
- GitHub repo
- Category
- Development
- Topic
- IoT & Hardware
- Open source
- Yes
- Platforms
- Linux, macOS, Windows, Command line, Code library, Self-hosted
- For
- IoT developers, smart home enthusiasts, privacy-focused developers, and edge AI researchers.
- Alternative to
- DensePose, MediaPipe
- Tags
- wifi sensing, csi, pose estimation, edge ai, esp32, home assistant, privacy
Why it was saved
- @tom_doerr: Privacy-first pose detection (no cameras), real-time processing (sub-50ms latency, 30 FPS), multi-person tracking of up to 10 individuals. MIT licensed, pip installable (pypi v1.1.0), Docker ready.
Related
Browser fingerprinting demonstration that tracks visitors across sessions via favicon caching
Updated