Low-Power Person Detection for UAV Search and Rescue Operations

This repository contains the code and documentation for the Low-Power Person Detection for UAV Search and Rescue Operations project in TinyML (Grade A) for the Stockholm University by Gabriele Pattarozzi, Timon Coucke and Jakub Schwenkbeck

For a quick idea what we did, take a look at this demo video:

tinyml_demo.mp4

Repository Structure

This repository is organized as follows:

• documents/ - project plan and report
• output/ - benchmarking results and plots
• models/ - final trained models and exported artifacts
• src/ - source code:
  • src/optimizing/ - model optimization scripts
  • src/training/ - training code and dataset preparation
  • src/visualization/ - plotting and visualization utilities
  • src/runtime/ - inference/runtime code for deployment on
• webui/ - code for a python based webui showcasing our key features including interference and benchmarking
• notebooks/pipeline.ipynb - comprehensive pipeline walkthrough (model loading → training → optimization → benchmarking).

Instructions

Model Optimization

Create a Python (3.11) virtual environment:

python3 -m venv opt_venv

Activate it:

source opt_venv/bin/activate

Install requirements:

pip install -r src/optimizing/requirements.txt

Model Runtime

Create a Python (3.11) virtual environment that includes the system site packages for picamera2 support:

python3 -m venv --system-site-packages venv

Activate it:

source venv/bin/activate

Install requirements and force reinstall for incompatible system site packages:

pip install -r requirements.txt