CrowdFace

• 

  Logo

• 

  Flowchart Diagram

• 

  Blake recognized on a LIVE Twitch stream

Inspiration

90,000 people are missing in the United States at any given time. Meanwhile, over 15 million are viewing live streams daily. Our mission is to engage the untapped crowdsourcing power of live stream viewers to spot missing children and catch wanted criminals.

What it does

Allows twitch chat users to verify by polling whether a face detected in the background of an IRL stream is in fact a match of a missing person or known criminal from our database.

How we built it

We mapped out the flow and split it into three engineering pieces. While the two other teammates worked on everything else!

Challenges we faced

Team dynamics; working 100% remote; time zone differences

Accomplishments we're proud of

Simultaneously generating more tips and better tips for set conditions for law enforcement to have increasing success in missing persons and trafficking cases

What we learned

How Twitch Extensions, OBS, and multiple other libraries work. We've never built anything like this before.

What's next for CrowdFace

Expand database of faces to detect Improve frequency of detection for static video streams Automate tip submission form tailored to agency preference

Built With

• cropper
• dlib
• docker
• face-recognition
• ffmpeg
• flask
• github
• jquery
• obs
• opencv
• pickle
• sklearn
• slack
• streamlink
• svm
• twilio
• twitch
• twitchdeveloperrig
• vote-now
• webex
• whereby
• zoom