EMMA-SC-SEP-v1¶

v1 · proposed-contract

Task. E-SC-SEP - see Tasks. What you predict. separated emitters and masks. Target domain. SPECTRUM - dense multi-emitter scene parsing for spectrum monitoring and electronic warfare.

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¶

Per-emitter separation masks.

Splits¶

train and dev are public and regenerable; holdout is private and scored by prediction submission. Sample count to be pinned at v1.

Get it¶

emma download --dataset EMMA-SC-SEP-v1 --split dev --out ./scenes

Metric¶

SISDR (scale-invariant signal-to-distortion ratio, in dB). See Metrics.

What this tests¶

Signal separation recovers overlapping emitters from a mixed radio-frequency (RF) scene, scored by scale-invariant SDR (signal-to-distortion ratio).

How it is made¶

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