WiFi DensePose: Track Every Move Without Cameras—Just WiFi Signals

This open-source project achieves something both unsettling and impressive: real-time indoor human pose tracking using nothing but ordinary WiFi router signals—no cameras, no sensors, no special hardware. It leverages WiFi Channel State Information (CSI), using neural networks to convert signal fluctuations into human body keypoint coordinates for DensePose-level pose estimation.

The specs are hardcore: sub-50ms latency, 30 FPS pose estimation, simultaneous tracking of up to 10 people. The Rust port is even more terrifying—810x faster than Python for the full pipeline, hitting 54,000 FPS throughput. It even includes a disaster response module (WiFi-Mat) that can detect vital signs through 5 meters of rubble with automatic triage classification.

The project claims to be "privacy-first," but consider this: your home WiFi router could theoretically become a 24/7 human behavior surveillance system. This project simultaneously showcases the brilliance and horror of technology—you might not even know WiFi signals are "watching" you.

Overview

WiFi DensePose (InvisPose) is a production-ready WiFi-based human pose estimation system. The core principle leverages WiFi router Channel State Information (CSI)—as WiFi signals propagate through space, human movement alters signal amplitude and phase. The system uses neural networks to parse these subtle changes into human body keypoint coordinates.

No cameras, no wearables, no special hardware—just ordinary WiFi routers.

Core Capabilities

  • **Real-time Pose Estimation**: <50ms latency, 30 FPS
  • **Multi-person Tracking**: Up to 10 simultaneous individuals
  • **Fall Detection**: Built-in analytics for fall detection and activity recognition
  • **Occupancy Monitoring**: Room occupancy counting
  • **WebSocket Streaming**: Real-time pose data push
  • **REST API**: Enterprise-grade with auth and rate limiting

Rust Port—Performance Monster

The Rust version delivers staggering improvements:

  • CSI preprocessing: 1000x faster than Python
  • Motion detection: 5400x faster
  • Full pipeline: 810x faster, 54,000 FPS throughput
  • Memory: 500MB → 100MB
  • WASM support for browser deployment
  • Mathematical precision validated: 0.000000 radians phase unwrapping error

WiFi-Mat Disaster Response

Specialized extension for search and rescue:

  • **Vital Signs**: Detect breathing (4-60 BPM) and heartbeat through 5m of rubble
  • **3D Localization**: Position estimation through debris
  • **Auto-Triage**: START triage classification (Immediate/Delayed/Minor/Deceased)
  • **Real-time Alerts**: Priority-based notifications
  • Use cases: earthquake, building collapse, avalanche, mine collapse, flood rescue

Installation

Supports Linux/macOS/Windows. Min 4GB RAM, 8GB+ recommended. GPU optional but recommended (NVIDIA CUDA).

Privacy & Ethics

The project claims "privacy-first" since no cameras are involved. But consider the flip side: precisely because no cameras are needed, this surveillance is more covert and harder to detect. Your neighbor, your landlord, anyone with router access could theoretically track your every move at home without your knowledge. This project is simultaneously a tech breakthrough and a security warning.

In-Depth Analysis and Industry Outlook

From a broader perspective, this development reflects the accelerating trend of AI technology transitioning from laboratories to industrial applications. Industry analysts widely agree that 2026 will be a pivotal year for AI commercialization. On the technical front, large model inference efficiency continues to improve while deployment costs decline, enabling more SMEs to access advanced AI capabilities. On the market front, enterprise expectations for AI investment returns are shifting from long-term strategic value to short-term quantifiable gains.