EMMA-S2T-CAP-v2¶
v2 · proposed-contract
Warning
Depends on an unshipped rfgen feature. The rfgen Phase 2 signal-to-text caption annotator is not yet shipped. This dataset cannot be regenerated until the pinned rfgen release lands; the contract is proposed, not final.
Task. E-S2T-CAP - see Tasks.
What you predict. a caption describing the scene.
Target domain. FRONTIER - radio-frequency (RF) understanding research.
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¶
Text caption.
Splits¶
train and dev are public and regenerable; holdout is private and scored by prediction submission. Sample count to be pinned at v2.
Get it¶
emma download --dataset EMMA-S2T-CAP-v2 --split dev --out ./scenes
Metric¶
BLEU (Bilingual Evaluation Understudy; n-gram precision), METEOR (Metric for Evaluation of Translation with Explicit ORdering; synonym and stemming match), and CIDEr (Consensus-based Image Description Evaluation; consensus captioning). See Metrics.
What this tests¶
Captioning tests whether a backbone can describe an RF scene in natural language, a task no community does at scale today.
How it is made¶
Generated by rfgen. The exact configuration (commit, emitters, channel, label extraction, seeds) is in the developer recipe.
See Also¶
Datasets: the full catalog.
Tasks: the task contract.
Developer recipe: rfgen configuration.