RuView

Overview

RuView is a WiFi sensing system that turns ordinary WiFi channel state information into real-time spatial intelligence: room occupancy, 17-point body pose, breathing and heart rate, and multi-person tracking, all without cameras, microphones, or wearables. It runs on the Cognitum stack with a Rust-based sensing pipeline and Ed25519-signed WASM edge modules, and can fuse its radio view with radar, lidar, depth cameras, thermal, acoustic, and BLE/UWB feeds into a single point cloud. The project is open source under the MIT license and is also delivered through Cognitum hardware tiers, from a low-cost ESP32-S3 node to a Seed and a v0 Appliance.

Key Features

• 17-point COCO body pose reconstructed from WiFi CSI, reported at 92.9% PCK@20 in upstream benchmarks
• Contactless vital signs, covering breathing at 6-30 BPM and heart rate at 40-120 BPM without a wearable
• Multi-person tracking of 3-5 or more people per access point, with persistent identity, pose, and vitals
• Through-wall sensing using Fresnel zones and multipath, including detection through 30 cm of concrete
• More than 60 signed WASM edge modules, each 5-30 KB, covering health, safety, building automation, analytics, and interaction
• Fusion runtime that merges WiFi CSI with radar, lidar, depth, thermal, acoustic, satellite, and BLE/UWB/Zigbee feeds

Use Cases

• Elderly care and sleep monitoring, including fall detection and contactless breathing and heart rate during sleep
• Retail and hospitality analytics such as queue length, dwell heatmaps, table turnover, and occupancy counts without cameras
• Smart building automation where room-level occupancy drives HVAC setpoints and lighting
• Search and rescue, where the WiFi-Mat kit detects vital signs through concrete and provides 3D survivor localization

Getting Started

• Pull and run the Docker image ruvnet/wifi-densepose:latest, publishing port 3000, then check the /health endpoint.
• Use the CLI with a simulator via ruview --source simulate, or connect an ESP32-S3 node over a serial port with ruview --source esp32.
• Query sensing state over REST at endpoints such as /api/v1/sensing/latest, /api/v1/vital-signs, and /api/v1/pose/current, or subscribe to the WebSocket stream on port 3001.

Deployment & Requirements

• The Docker container exposes port 3000 for REST requests and port 3001 for the live WebSocket stream.
• Radio capture requires ESP32-S3 sensor nodes; the recommended Seed adds persistent local memory and Ed25519 hardware-key attestation.
• On-device inference runs on a Cognitum Seed or a v0 Appliance; the v0 Appliance is a dedicated hub for 3-6 ESP32-S3 nodes with a private REST and WebSocket API on the LAN and can run air-gapped.
• Cloud aggregation is optional; frames stay in the room unless a cloud destination is explicitly configured.

Before You Adopt

• License: MIT. Review its terms before using, modifying, or distributing the project.
• Health and vitals modules are released as developer previews, and regulated clinical uses require validation beyond the stock repository.
• Field accuracy depends on the radio-frequency environment, with per-room adaptation taking up to about 30 seconds.
• Published performance figures come from simulations and prototypes and may not match final production hardware.
• Compliance statements describe architectural posture and design intent; regulated production deployments need customer-specific validation.