Insights > Customer Stories

Combating Child Abuse with Machine Learning

When the nonprofit organization Project VIC International wanted to create a proof-of-concept of its innovative solution for identifying specific items in online child abuse imagery, they partnered with Redapt to build it.




In a nutshell ...

1
The problem

The nonprofit organization Project VIC International needed a proof-of-concept for its innovative imagery tracking model that assists law enforcement in finding and rescuing abused children. The goal of the technology is to make it possible for agencies to find victims faster.

2
The solution

Redapt partnered with the organization and Microsoft to deploy its database of imagery in Azure for Project VIC International’s models to automatically sanitate and scan unlawful images via machine learning.




3
The outcome

The proof-of-concept project was a success, allowing Project VIC International to continue and refine its work as it sought out further funding for its important mission.




Moment of clarity

While Project VIC International initially partnered with Microsoft to build the identification models, after reviewing the models already put together by the organization, Redapt determined that much of the heavy lifting had already been completed. Instead, our focus would be on implementing and optimizing the technology in Azure.

What the company needed

The nonprofit organization Project VIC International is dedicated to helping law enforcement agencies locate children that are victims of child abuse using online imagery and videos. The organization’s focus is on using technologies to identify images such as logos and consumer items visible in photos to assist in narrowing the search for children being abused.

To achieve this, Project VIC International amassed a database of imagery from around the world and is developing machine learning models that can be used by law enforcement to match items visible in child abuse imagery to the images in the organization’s database.

With funding from Microsoft, Project VIC International partnered with Redapt to implement a proof-of-concept project to prove the organization’s models could make a successful identification.




Our recommendation

Working with Project VIC International, Redapt deployed two machine learning models—one to extract potential clues from images (known as “chips”) and the other to match those images with identical ones in random photos.

In addition, we provisioned a Linux virtual machine in Azure and created a python/javascript web application that allowed for child abuse images to be uploaded and sanitized before the models were run.




The end result

Once implemented, Project VIC International’s models were able to identify specific logos in test imagery. This allows the organization to further refine and build upon its models while it seeks out additional funding, with the end goal of one day providing law enforcement agencies with a critical tool for combating child abuse online.

Contact us to gain a little more clarity

We love a challenge. Tell us about your needs, goals, and the obstacles in your way. We’ll help you find solutions to get past them.

Get in touch directly:

results@redapt.com
(425) 882-0400