Projects · 06
WiFi Vision
By Haider Ali ·
WiFi Vision, by Haider Ali, tracks several people moving through a room from WiFi signals alone, using a mesh of ESP32 nodes. No cameras, no wearables.
Tracks several people moving through a room from WiFi signals alone, using a mesh of ESP32 nodes. No cameras, no wearables.
Jun–Aug 2026 · shipped
WiFi Vision tracks several people moving through a room from WiFi signals alone: no cameras, no wearables. In a six-node simulation it tracks two walkers to about 0.2 m.
Key facts
- Device-free, multi-person tracking
- ~0.2 m RMSE in simulation
- 124 commits
How it works
- Sense: ESP-IDF firmware on every ESP32 node captures WiFi channel state information.
- Stream: the nodes send it over ESP-NOW to a TypeScript host.
- Locate: per-link features feed radio tomographic imaging, which solves for position.
- Track: a multi-person Kalman tracker drives live 2D and 3D dashboards.
Details
- Honest confidence: when it can't separate people, it says “1+?”.
- One firmware image for every node; a web flasher sets up the whole fleet.
- Replay mode needs no hardware.
Stack
- c / esp-idf
- esp32-c6
- esp-now
- typescript
- next.js
- websocket
- three.js
Private repo: the source isn’t public.
#embedded #sensing #esp32