TraffickCam: Crowdsourced and Computer Vision Based Approaches to Fighting Sex Trafficking

Abstract

According to a 2016 study by researchers at the University of New Hampshire, over sixty percent of child sex trafficking survivors were at one point advertised online [13]. These advertisements often include photos of the victim posed provocatively in a hotel room. It is imperative that law enforcement be able to quickly identify where these photos were taken to determine where a trafficker moves their victims. In previous work, we proposed a system to crowdsource the collection of hotel room photos that could be searched using different local feature and image descriptors. In this work, we present the fully realized crowd-sourcing platform, called TraffickCam, report on its usage by the public, and present a production system for fast national search by image, based on features extracted from a neural network trained explicitly for this purpose.