Inspiration

ThreatHunter was inspired by ongoing human trafficking and child abduction issues where there are 40.3 million victims of human trafficking globally, and 25% are children. In the United States, a child is missing or abducted every 40 seconds. The National Center for Missing & Exploited Children states that of more than the 23,000 runaways reported in 2018, one in seven were likely victims of child sex trafficking.

Amber alerts have been implemented as a child abduction alert system to send text messages to phones in a surrounding area asking for the public to help find abducted children.

However, once receiving the text message, people are unlikely to remember or memorize the actual license plate number. With ThreatHunter, we are able to actively engage the public to improve searching and detecting by providing them a platform to livestream their surroundings to a server, which will extract license plate numbers from the footage and identify if they match the number from the Amber alert.

What it does

We provide an app that people can use whenever an Amber alert gets dispatched in the area to livestream their surroundings (walking, mounted on a cardash, etc). Our algorithm will cross-reference the license plate provided from the Amber alert with the livestreams provided from the distinct area that got the alert by using object detection. Once a match is found, the coordinates of the detecting phone gets sent to law enforcement where they can track the location of the offender.

How we built it

  • Community application was built using Android studio.
  • Backend function trigger using with IBM Cloud Functions
  • AI custom-trained model for object detection using IBM Cloud Annotations
  • Detect text from image using AWS Rekcognition
  • Twilio for SMS automation

Challenges we ran into

  • Having large amounts of data to improve the model accuracy

Accomplishments that we're proud of

  • Extracting the license plate numbers from an image to text using AWS Rekcognition
  • Training the IBM Watson model with our footage
  • Serverless function

What we learned

  • Using IBM Watson to train a model
  • Creating an Android application
  • AWS Rekcognition for text detection
  • Twilio for SMS automation

What's next for ThreatHunter

We hope to create the law enforcement side of the application with real time updates and tracking of the offender's location using Google Maps. We also would like to train the Watson Machine Learning model with more data for better accuracy in detecting license plates.

Share this project:
×

Updates