Abstract

Investigations of sex trafficking sometimes have access to photographs ofvictims in hotel rooms. These images directly link victims to places, which canhelp verify where victims have been trafficked or where traffickers mightoperate in the future. Current machine learning approaches give promisingresults in image search to find the matching hotel. This paper exploresapproaches to make this end-to-end system better support government and lawenforcement requirements, including improved performance, visualizationapproaches that explain what parts of the image led to a match, andinfrastructure to support exporting the results of a query.

 

Quick Read (beta)