Inspiration
The internet is the largest platform for communicating and meeting people around the world. Websites like Omegle and Chat Roulette make incredibly easy to introduced to new individuals with new experiences. However, it leaves many people vulnerable to unwarranted harassment and unsavory individuals
Our goal with SafeFace is to use active image and video recognition using Clarifai to filter what individuals experience, so they can spend more time interacting with others with like affinities.
What is it?
SafeFace is a web based anonymous chat client like Chat Roulette or Omegle which allows you to filter what people are sharing using active visual recognition before and during a chat.
For example, by putting 'nsfw' our server will run a visual check on the user's video feed before connecting with you, to determine if what they are displaying is deemed safe for work, by a predefined prediction threshold. This is not limited to not safe for work content and can span out across all of Clarifai's concept tags, as input by the user.
Construction
Our web server is written using Node.js and Socket.io to handle incoming traffic and properly managing connecting clients together with Clarifai interpreting individuals video feed.
Our server takes a snapshot of video feed before connecting two clients encoding it into Base64 which is then interpreted by the Clarafai API, and compares all tags the user has selected to Clarafai's predicted tags within a valid threshold. This can apply to a large array of topics, making it capable of filtering lots of specific information.
If the user is displaying content one wishes not to see, it terminates the connection before before or as soon as one of the users filters is detected.
Challenges
Initially creating the double video feed, implemented a proper connection queue for creating lobbies was the hardest feat. Then, once we had the webpage up and running the biggest difficulty was interpreting the users video feed before a connection was made.
Next step was formatting the web page with a smooth HTML layout, with very mild success we achieved that.
Accomplishments
Everything we've done with the project we are incredibly proud of, with the little knowledge we had of JavaScript we're very proud of what we were able to develop.
What we learned
We learned lots of JavaScript, server management, good use of the Clarifai API, and most importantly that it's not necessary for individuals to have to be exposed to unwarranted harassment by individuals who are just trying to invoke a negative reaction.
What's next?
We plan to fleshout the layout of the webpage as well as improving our servers connection protocols to ensure smooth lobby making, and matching those with proper filters more efficiently.
Log in or sign up for Devpost to join the conversation.