Top AI Product

So here’s something that sounds like it belongs in a sci-fi movie but is actually sitting on [GitHub right now](https://github.com/ruvnet/wifi-densepose) with nearly 7,000 stars. wifi-densepose is a production-ready system that uses your regular WiFi signals to track human body poses in real time — through walls, without a single camera involved.

The concept comes from a [Carnegie Mellon research paper](https://arxiv.org/abs/2301.00250) called “DensePose From WiFi,” which showed that WiFi signals bouncing off human bodies carry enough information to reconstruct full body poses. The original research was impressive but mostly academic. What wifi-densepose does is take that idea and turn it into something you can actually deploy. We’re talking a proper Python package, Docker support, production APIs with auth and rate limiting — the whole deal.

Here’s how it works in plain terms: WiFi routers constantly send out signals, and when those signals hit a person, they bounce back differently depending on the body’s shape and position. The system captures this Channel State Information (CSI) and feeds it through a neural network that maps it to dense human pose coordinates across 24 body regions. The result? Full-body tracking at 30 FPS with under 50ms latency, and it can handle up to 10 people at once.

What makes this genuinely interesting to me is the privacy angle. Camera-based tracking systems always come with that uncomfortable feeling of being watched. This approach sidesteps that entirely — there’s no visual data being captured, period. That makes it way more palatable for things like elderly care monitoring, home fitness tracking, or smart home automation where people don’t want a camera staring at them 24/7.

The project shot up on the [GitHub Trending Python weekly chart](https://github.com/trending/python?since=weekly) recently, gaining over 1,100 stars in a single week. It’s not hard to see why — the idea of repurposing cheap mesh routers into a full body tracking system hits that sweet spot between “that’s wild” and “wait, I could actually use this.” If you’ve got a few WiFi routers lying around and some Python chops, it’s worth checking out the [repo](https://github.com/ruvnet/wifi-densepose) and poking around.