Who RuView is for#
Healthcare facility
Monitors patient breathing and movement overnight without installing cameras in patient rooms, avoiding imaging privacy regulations.
Building management team
Counts occupancy across meeting rooms using existing WiFi infrastructure and adjusts HVAC automatically.
Developer / elderly care app
Builds a fall-detection system for elderly care that alerts a caregiver when no motion is detected for an extended period.
Retail store
Measures dwell time and foot traffic patterns without video surveillance to comply with privacy regulations.
Researcher
Uses the Docker image to prototype and validate signal-processing pipelines before committing to hardware deployment.
The problem it solves#
Occupancy sensors require line-of-sight, cameras create privacy concerns and regulatory friction under GDPR and HIPAA, and wearable health monitors only work when someone remembers to put them on. Building a system that can detect presence, vital signs, and activity across a space typically means choosing between invasive video surveillance, expensive dedicated radar hardware, or passive infrared sensors that miss stationary occupants entirely.
How it solves it#
Through-wall presence and vital signs
Detects presence, breathing rate (6-30 BPM), and heart rate (40-120 BPM) through walls using WiFi Channel State Information (CSI)
WiFi-based pose estimation, no cameras
Estimates 17 COCO body keypoints (pose skeleton) from WiFi signals using the WiFlow neural architecture, no cameras needed
Full edge deployment, no cloud
Runs entirely on edge hardware: an ESP32-S3 mesh ($9 per node) with no cloud, no internet required after setup
Multi-frequency mesh scanning
Multi-frequency mesh scanning across 6 WiFi channels, using nearby access points as passive radar illuminators at no extra cost
Full system BOM around $140
Full system BOM around $140 including a Cognitum Seed for persistent vector storage and cryptographic attestation
Docker image for hardware-free evaluation
Docker image available for simulation and pipeline evaluation without any hardware
Strengths and trade-offs#
Strengths
- No cameras, no GDPR/HIPAA imaging frictionNo cameras means no GDPR video or HIPAA imaging compliance burden by design, making deployments in healthcare and residential settings straightforward
- Self-adapts in 30 seconds, no calibrationSelf-adapts to each room environment in under 30 seconds using spiking neural networks, no manual calibration or training data collection required
- Low hardware cost, reuses existing WiFiHardware costs are genuinely low: a basic sensing node is a $9 ESP32-S3 and most deployments can reuse existing WiFi infrastructure
- MIT licensed with 1,463+ passing testsMIT licensed with 1,463+ passing tests and active development
Trade-offs
- -Camera-free pose accuracy still improvingCamera-supervised pose accuracy (PCK@20 target of 35%+) is still in the evaluation phase; current camera-free accuracy is limited to approximately 2.5% PCK@20 with proxy labels
- -Requires ESP32-S3 specificallyESP32-C3 and original ESP32 are not supported; requires ESP32-S3 specifically due to DSP requirements
- -Multi-node recommended for larger spacesSingle-node deployments have limited spatial resolution; two or more nodes are recommended for accurate person counting in larger spaces
RuView vs alternatives#
RuView is an MIT-licensed open-source edge-AI alternative to AWS IoT Greengrass, Amazon Bedrock Agents, and Datadog edge monitoring for teams experimenting with local sensing, WiFi CSI, and real-time physical-space analytics.
vs AWS IoT Greengrass: AWS IoT Greengrass helps teams build, deploy, and manage device software, run local processing, and perform machine-learning inference across edge devices. RuView is narrower and more experimental: it focuses on WiFi DensePose, ESP32 CSI data, pose estimation, vital signs, WebSocket streams, and Docker-based demos. RuView is better for research teams that want open sensing pipelines they can inspect and modify. Greengrass is still better for production fleet management, AWS integration, and managed deployment controls.
vs Amazon Bedrock Agents: Bedrock Agents orchestrate foundation models, data sources, APIs, memory, monitoring, encryption, and permissions for hosted generative-AI apps. RuView is not a hosted agent platform. It is a sensing and edge-processing stack that can expose live sensor data and local analytics to applications. Bedrock Agents are still better when the main job is building managed customer-facing AI workflows.
vs Datadog: Datadog provides top-to-bottom edge fleet monitoring, dashboards, logs, and drill-down troubleshooting across device dimensions. RuView is better when teams need local signal processing and visual sensing output, not a managed observability SaaS. Datadog is still better for production alerts, operational reporting, and cross-fleet monitoring.
Install and self-host#
docker run -p 3000:3000 ruvnet/wifi-densepose:latestWhat it's built on#
- Languages
- CJavaScriptPythonRustTypeScript
- Frameworks
- React
FAQ#
Does this require special WiFi hardware?
For full sensing (presence, vital signs, pose), you need an ESP32-S3 ($9) for CSI capture. The Docker image runs on simulated data with any computer. Consumer WiFi laptops provide only coarse RSSI-based presence detection.
Does it work through multiple walls?
Through-wall detection works up to about 5 meters using Fresnel zone geometry and multipath modeling. Accuracy degrades with distance and wall density; denser building materials reduce effective range.
Is data sent to the cloud?
No. RuView processes everything on edge hardware. The Cognitum Seed adds persistent vector storage and kNN search locally. Internet access is not required after initial firmware flashing.
Similar open-source tools#
Coroot
Instant observability with no-code setup.
RuFlo
Deploy intelligent AI agents with ease.
Sentry
Real-time error tracking with performance monitoring and traces
ThingsBoard
Open source IoT platform for device management and dashboards
ClawTrace
Visualize agent execution trees and track token costs per step
Uptime Kuma
Track uptime for websites and APIs with 90+ alert integrations