Datasets for Radio Deep Learning

At the Seventh Annual GNU Radio Conference (GRCon17), for which DeepSig was a sponsor, DARPA MTO Program Manager Paul Tilghman ran a workshop about RF Datasets where he called for the creation of an open & publicly available dataset of RF recordings to enable open data science, acknowledging the impact that DeepSig's previously published datasets have had in this field. Before the workshop concluded, Paul formed a working group of industry experts to lead the creation of just such a dataset with the goal of having it ready by GRCon18.

Ben Hilburn, the Director of Engineering at DeepSig and President of the GNU Radio Foundation, is one of the four members of this working group. Ben is the lead developer of the Signal Metadata Format (SigMF) specification, which the working group will be using in the creation of its dataset. Per the Abstract of the SigMF Specification, "The Signal Metadata Format (SigMF) specifies a way to describe sets of recorded digital signal samples with metadata written in JSON. SigMF can be used to describe general information about a collection of samples, the characteristics of the system that generated the samples, and features of the signal itself."

The working group is meeting regularly and the SigMF specification process continues to gain momentum. If you are interested in getting involved with the design of SigMF, check out the "Contributing" section of the SigMF documentation.

The slides from Paul Tilghman's workshop are available on the GRCon17 Program page here. For all other inquiries related to the working group or SigMF, you can contact us at