open-source · WiFi CSI sensing
See a room
through WiFi.
No camera.
RuView reads the way ordinary WiFi signals bend and scatter around a body, and turns that into presence, movement, and breathing signals — processed at the edge, published as open source.
MIT licensed · ESP32-class hardware · camera-free by design
01 — try it now
Run the browser demo
The interface below is the actual RuView demo, served from the project's GitHub Pages build. It runs entirely in your browser — nothing is uploaded, because there's no camera feed to send.
Frame not loading? Your browser may be blocking embedded content. Open the demo in a new tab ↗
- CSISignal input, not video
- $9+Cost of a sensing node
- EdgeProcessed on-device
- 17ptPose keypoints, research path
- VitalsBreathing & pulse trends
- MITFully open source
02 — the pipeline
From radio static to room awareness
Four stages turn a raw WiFi channel reading into something a program can act on.
-
01
Capture the channel
A sensing node — commonly an ESP32-S3 — samples Channel State Information from the WiFi link: how the signal's amplitude and phase shift on every subcarrier.
-
02
Isolate the movement
Static reflections from walls and furniture are filtered out, leaving the smaller fluctuations caused by a body moving, breathing, or shifting weight.
-
03
Run inference locally
A lightweight model turns the filtered signal into structured output — presence, coarse pose, or a breathing-rate estimate — without sending anything off-device.
-
04
Expose the result
The demo renders that output live in the browser; in a real deployment it's the signal your app, dashboard, or alert logic would subscribe to.
03 — learn
What is WiFi CSI sensing?
A short explainer for anyone landing on this page without a signal-processing background.
Every WiFi connection already measures something extra: how the radio signal itself changed on its way from router to device. That measurement is called Channel State Information, or CSI — a description of amplitude and phase across each WiFi subcarrier, refreshed many times a second.
Walls, furniture, and other fixed objects create a stable CSI pattern. When a person moves through that same space, their body absorbs and reflects part of the signal, and the pattern shifts in small, structured ways. Strip out the stable background and what's left is a trace of motion — footsteps, a hand raised, a chest rising and falling while someone sleeps.
That's the whole premise behind RuView: reuse a signal that's already there, filter it down to the part that moves, and let a small model turn that into presence, activity, or a breathing-rate estimate. No image is ever formed, because there's nothing image-shaped in the data to begin with.
- CSI
- Channel State Information — how a WiFi signal's amplitude and phase change across subcarriers as it travels through a space.
- RSSI
- Received Signal Strength Indicator — a single coarse number most WiFi chips expose. Useful for rough distance, too coarse for detailed sensing.
- Edge processing
- Running inference on the sensing device itself rather than sending raw data to a server.
- WiFi DensePose
- A research direction that maps WiFi signal changes to estimated human body position, similar in spirit to camera-based pose estimation.
04 — what it does
Capabilities
CSI-based sensing
Reads amplitude and phase across WiFi subcarriers to infer occupancy and movement in a space.
Camera-free by default
Built for bedrooms, bathrooms, care facilities, and anywhere a lens is unwelcome or impractical.
Vital-sign trends
Surfaces contactless breathing and heart-rate patterns from small variations in reflected signal.
Edge-first processing
Sensor nodes pair with local inference, so sensitive signal data stays close to where it's captured.
Room-level intelligence
Supports research into occupancy, activity type, room fingerprinting, and fall-risk indicators.
Open development
Source, firmware notes, and model references are public — issues and contributions happen on GitHub.
05 — where it fits
Use cases
Elder care & health research
Prototype contactless sleep tracking, inactivity alerts, and fall-risk signals without asking anyone to wear a device.
Smart building sensing
Study meeting-room occupancy, HVAC triggers, and space utilization without installing cameras.
WiFi sensing research
Use RuView as a starting point for pose-estimation, signal-fusion, and edge-model experiments on low-cost hardware.
Before you rely on this
- Treat every output — especially vitals, sleep, and fall signals — as an experimental research estimate, not a validated medical reading.
- Ordinary laptop WiFi chipsets typically expose only coarse signal strength (RSSI); real CSI sensing needs compatible hardware such as ESP32-S3-class nodes.
- Accuracy improves with multiple sensing nodes and a calibrated environment — a single device in an unfamiliar room will be noisier.
- Radio-based sensing doesn't record video, but it can still reveal occupancy and activity patterns — plan for that the same way you would for any sensing system.
06 — questions
Frequently asked
What is RuView?
An open-source project exploring WiFi Channel State Information as a sensing medium — for presence detection, movement, breathing and heart-rate trends, and camera-free spatial awareness.
Can I try it without any hardware?
Yes — the embedded demo above runs in your browser so you can explore the interface before setting up any sensing hardware.
What hardware does real sensing need?
Meaningful CSI capture needs a chipset that exposes it — ESP32-S3-class sensor nodes are the common low-cost path. Most laptop WiFi cards only report signal strength, which isn't enough on its own.
Does it use a camera or a wearable?
No. Sensing is based entirely on changes in radio signal, so there's no lens and nothing to wear.
Is this ready for medical or safety use?
No — it's active open-source research. Any health, safety, or security use needs independent validation first.
How is WiFi sensing different from radar or lidar?
Radar and lidar use dedicated transmitters built for ranging. WiFi sensing repurposes signals already broadcast by routers and devices you likely own, trading purpose-built precision for near-zero extra hardware cost.
Does RuView send my data anywhere?
The design goal is edge-first processing, meaning inference runs on the local device rather than a cloud service. Always check the current firmware and app configuration for your specific setup.
Where's the source code?
On GitHub, including firmware notes, setup steps, and model references — linked throughout this page.