ruvnet · 2026-04-16 · notable
RuView — WiFi DensePose: 17-Point Body Pose Through Walls, No Camera Required
RuView uses $9 ESP32 WiFi nodes and a 1.8M-parameter graph transformer to estimate 17-point body pose, breathing rate, and heart rate through walls in real time — no camera, no wearable. 47.9k GitHub stars, trending #2 today.
Real-time body pose and vital sign monitoring through walls using commodity WiFi — no camera, no wearable, $9 hardware.
Key specs
| GitHub stars | 47,900+ |
|---|---|
| Inference latency | 0.012ms |
| Pose accuracy (pck@20) | 92.9% |
| Processing speed | 11,665 fps |
| Model size (quantized) | 881 KB |
What is it?
RuView is an open-source edge-AI system that turns cheap ESP32-S3 WiFi nodes into ambient human sensors. Point two nodes at a room and it reconstructs 17-point body pose, breathing rate, heart rate, and multi-person presence in real time — through walls, in complete darkness, with no camera. The latest release is v0.6.1-esp32, April 16, 2026. 47.9k GitHub stars.
How does it work?
Each ESP32-S3 node samples WiFi Channel State Information (CSI) at 20 Hz. The signal passes through SpotFi spatial fingerprinting and a Hampel outlier filter, then feeds a graph transformer with cross-attention: 1.8M parameters quantized to 881 KB (4-bit form: 8 KB). The inference stack is written in Rust — 810x faster than the original Python version — hitting 11,665 frames/second at 0.012ms latency. Micro-LoRA with Elastic Weight Consolidation enables on-device personalization without catastrophic forgetting.
Why does it matter?
Most ambient sensing needs cameras (privacy risk) or expensive radar hardware. RuView reaches 92.9% PCK@20 pose accuracy and 100% presence detection using $9 commodity WiFi chips. Whether you're building elder-care monitoring or analyzing surveillance attack surfaces, this is the clearest working demonstration of what a trained edge model on WiFi CSI can actually do — and it surfaces a real gap in privacy law, since CSI-based sensing doesn't trigger existing camera-focused regulations.
Who is it for?
Edge ML engineers, smart-home developers, security researchers, and privacy advocates.
Try it
https://github.com/ruvnet/RuView