AI Deck

RuView — A Rust open-source platform that turns commodity WiFi signals into camera-free pose, breathing, and presence detection

RuView is an open-source software platform that repurposes the WiFi signals already filling your room as a sensor. The radio waves traveling between an access point and its clients are perturbed slightly every time a person moves or breathes. By feeding the record of those perturbations — CSI (Channel State Information) — into machine learning models, RuView infers occupancy, headcount, body pose, breathing rate, heart rate, and falls without a camera or a wearable. It is developed by Canadian engineer Reuven Cohen (ruvnet), written in Rust, and released under the MIT license. It is published on GitHub, where it has gathered more than 90,000 stars.

Key Features

  • Pose and vital signs from WiFi CSI: Supports breathing rate (6–30 breaths per minute), heart rate (40–120 beats per minute), 17-keypoint pose estimation, fall detection (responding in under 200 milliseconds), multi-person counting, and sleep stage classification
  • Pretrained models you can use as-is: wifi-densepose-pretrained (a CSI encoder) and wifi-densepose-mmfi-pose (17-keypoint pose estimation) are distributed on HuggingFace, so you can try the system without collecting your own data. Quantized variants ranging from a few kilobytes to a few dozen kilobytes are also available
  • Buildable from inexpensive off-the-shelf hardware: The basic setup places several ESP32-S3 nodes (about $9 each) or WiFi 6-capable ESP32-C6 boards in a mesh to collect CSI. Research NICs (Intel 5300, Atheros AR9580) are supported as well
  • Extensible through edge modules: More than 105 modules across 11 categories — health, security, building management, retail, industrial, research, and others — let you load only what you need
  • Smart home integration: Connects to Home Assistant, Apple Home, Google Home, Alexa, and Matter, so presence detection can be wired into automations you already run
  • Local-only processing for privacy: No video is ever captured, and processing stays on your own hardware. The project passes 1,463 tests and ships releases at a high cadence

Pricing

PlanCostWhat you get
SoftwareFree (MIT license)All functionality. Commercial use, modification, and redistribution are permitted
HardwareAt costRoughly $9 per ESP32-S3 node, $6–10 for an ESP32-C6. The preassembled “Cognitum Seed” bundle referenced by the project costs about $140

Pricing reflects information as of August 2026. Check the official repository for the latest details.

Pros & Cons

Pros

  • Captures presence, pose, and vital signs without installing a camera, so it works in places where you would rather not put a lens, such as near a bedroom or changing area
  • Radio waves pass through walls and furniture to a degree, so there are fewer blind spots than with a lens
  • Free under the MIT license: no software cost, and few restrictions on commercial use
  • Pretrained models are published, which removes the heavy up-front work of data collection
  • Because it runs on ESP32 boards, you can start experimenting for roughly ten dollars per node

⚠️ Cons

  • You must source, place, and power the CSI-capable hardware yourself. This is not an out-of-the-box product like a retail sensor
  • It assumes a Rust build environment and command-line familiarity, so a non-engineer will struggle to deploy it alone
  • Accuracy depends on the radio environment — room shape, furniture layout, nearby WiFi devices — and needs tuning per installation
  • The published accuracy figures (such as 82.69% torso-PCK@20 for pose estimation) are research benchmark numbers, not a level of reliability comparable to a medical device
  • Development moves very fast, and configuration can change between versions

Comparison with Similar Services

CriterionRuViewmmWave presence sensorMonitoring cameraWearable device
Sensing methodExisting WiFi signals (CSI)Millimeter-wave radarVideoBody-worn sensor
CameraNot usedNot usedUsedNot used
Information capturedPresence, headcount, pose, breathing, heart rate, fallsPresence, motion, rough positionVideo itself, motion detectionHeart rate, sleep, step count
Needs to be wornNoNoNoYes
Ease of setupRequires a DIY build; difficultBuy the product and install itBuy the product and install itBuy it and wear it
CostFree software plus ESP32 hardwareProduct priceProduct price plus cloud feesProduct price

Who Is It For

  • People who want to watch over a family member’s presence or detect falls without a camera, and who can accept the DIY effort involved
  • Students and researchers studying WiFi sensing and CSI
  • Anyone comfortable with ESP32 and DIY IoT who wants to strengthen Home Assistant automations with radio-based presence detection
  • Anyone who needs presence detection in a setting where video or cloud transmission must be avoided for privacy reasons
  • Developers looking for a permissively licensed WiFi sensing implementation they can embed in a commercial product

Summary

RuView is an open-source project that repurposes the WiFi signals already crossing your room as a new kind of sensor. Inferring presence, pose, and breathing without a camera or a worn device is genuinely appealing for monitoring and smart home automation. On the other hand, you have to handle everything yourself, from sourcing and placing hardware to building the Rust code, so it cannot be adopted the way you would buy a finished sensor. A realistic starting point is to gather a few ESP32-S3 boards and try presence detection with the pretrained models. It is also worth drawing a clear line around the vital sign figures: treat breathing and heart rate as reference values, not as a basis for medical judgment.

← Blog