Dataset Manager

The Dataset Manager is RIA Hub’s suite of tools for creating and validating RF machine learning datasets. It covers the full dataset lifecycle: generating synthetic labelled data from scratch, curating raw recordings into structured training sets, and inspecting the results for quality and balance.

Module	What it does
Synthesizing a Dataset	Generate labelled synthetic IQ datasets by configuring modulations, channel conditions, and SNR ranges — no hardware required
Curating a Dataset	Transform raw RF recordings into production-ready HDF5 datasets through a guided pipeline of slicing, quality filtering, and optional augmentation
Inspecting a Dataset	Analyse an existing dataset for class balance, per-class signal statistics, anomalies, and differences between versions

Choosing the right tool

Start with the Generator if you don’t have recordings yet and want to prototype a modulation-recognition model quickly. Synthetic datasets are fast to produce and fully reproducible.

Use the Curator when you have real-world recordings (e.g. from a Campaign Control run) and want to turn them into a training-ready dataset with quality filtering and metadata preserved.

Use the Inspector after generating or curating to verify the dataset looks as expected before committing it for training — catching class imbalance or signal anomalies at this stage is much cheaper than discovering them after a training run.