drau

Caches the Hugging Face dataset geronimobasso/drone-audio-detection-samples and provides tools for running acoustic drone-detection tests against external microphone systems.

Requirements

• Python ≥ 3.11
• PortAudio — installed automatically by make venv on Debian/Ubuntu; on macOS: brew install portaudio

Quickstart

Copy .env.example to .env and set HF_TOKEN if the dataset requires authentication.

cp .env.example .env
nano .env

Create a token at https://huggingface.co/settings/tokens, set and save:

HF_TOKEN=hf_****************

Then:

make venv          # create .venv, install dependencies, and install libportaudio2
make cache-data    # download and unpack the dataset (~6.6 GB)

Run make help for all available targets and examples.

run-session

Runs a calibrated detection test session and auto-analyses the results.

Parameter	Required	Description
SAMPLES	yes	Number of audio samples to play
MICS	yes (unless PLAY_SAMPLE=1)	Number of microphone systems under test
DIST_MAX	yes	Maximum simulated drone distance in metres
PLAY_SAMPLE	no	Play samples continuously without collecting results

# Full detection test session
make run-session SAMPLES=20 MICS=2 DIST_MAX=150

# Play-only preview — no mic input or CSV written
make run-session SAMPLES=20 DIST_MAX=100 PLAY_SAMPLE=1

Play-only mode uses the same 50/50 drone/non-drone selection and distance simulation as a full session, but auto-advances through samples without waiting for mic identefication input.