xView3 Challenge and 7th Place Submission

From September to November 2021 I led the TRSS team on the 2021 xView challenge problem from the Defense Innovation Unit. The problem was to identify ships from Synthetic Aperture Radar data.

I don’t care to walk through the piece again when there is a formal write up available here. Suffice it to say, we did well, using a modification of a traditional radar sensing technique and a relatively established CNN.

Thank you TRSS for sponsoring the work — it was a privilege.

Two other pieces of TRSS work from the same stretch: the anomix package and the hashtag community detection paper.

Linked from

Hashtag Community Detection on Social Networks

A paper on grouping hashtags into communities using metadata-derived entropy scores, applied to Twitter and Parler.

Anomix: Mixture Models for Anomaly Detection

An open source Python package for univariate anomaly detection with mixture models. Narrow, but it does its few jobs well.

About

Data scientist in Washington, D.C. Statistics, energy, bicycles.