EMMA-DRONE-v0.1¶

v0.1 · proposed-contract

Task. E-ID-DRONE - see Tasks. What you predict. whether a drone is present, and which model. Target domain. DEFENSE / SPECTRUM - defense and spectrum monitoring (every public drone radio-frequency (RF) dataset is single-antenna, so drone angle-of-arrival (AoA) is unevaluable elsewhere).

Input¶

Raw complex multi-antenna I/Q (in-phase and quadrature), shape (num_rx, 2, N) float32, with per-scene metadata. 8-element uniform linear array (ULA).

Labels¶

A drone-present flag plus a drone-model class.

Splits¶

train and dev are public and regenerable; holdout is private and scored by prediction submission. About 120,000 scenes.

Get it¶

emma download --dataset EMMA-DRONE-v0.1 --split dev --out ./scenes

Metric¶

AUC (area under the ROC curve). See Metrics.

What this tests¶

Every public drone-RF dataset is single-antenna, so multi-antenna drone detection and direction-finding are unevaluable in the existing literature.

How it is made¶

Generated by rfgen. The exact configuration (commit, emitters, channel, label extraction, seeds) is in the developer recipe.