This repository contains a webcam-based inference demo inspired by PoseFly, a 4-in-1 Optical Camera Communication (OCC) approach for swarming drones that integrates:
- Drone identification (active optical label via rolling-shutter OOK),
- On-site localization / pose parsing (distance, angle, speed estimation),
- Quick-link communication (LED-to-camera burst data),
- Lighting (LEDs as illumination). :contentReference[oaicite:0]{index=0}
PoseFly uses rolling-shutter cameras and simple LED hardware to support drone-to-drone interactions without relying on congested RF-only centralized control. :contentReference[oaicite:1]{index=1}
In this implementation, we run a webcam inference pipeline that can optionally enable/disable:
- 🛸 Drone Detection
- 🧭 Angle Estimation
- 📏 Distance Estimation
- 💡 LED ID / Optical Label ID
- 🎞️ Optional video saving
The original PoseFly paper reports strong prototype results (e.g., near 100% distance estimation accuracy up to 20 m, identification up to 12 m, angle/speed within 4 m, and ~5 Kbps quick-link throughput on prototypes).