Meme Relief

Inspiration

All of us being the spiciest of meme connoisseurs, wanted a way to stop ourselves from seeing overused memes while casually browsing the internet. Although the original basis for this project did not have practical applications, we found that this software tool could be utilized by teachers or employers to increase productivity in the workplaces and schools by preventing employees/students from seeing the majority of inappropriate images.

What it does

We specifically made the software to block all spongebob memes with a model made with microsoft custom vision. Microsoft custom vision works by prompting the user to submit pictures (either unlabeled or having the user label them his or herself), teaching itself which important concepts to use when classifying images, and finally tag images with successful classifcations.

Our software framework sits on top of this computer vision model and creates an array of all images on a web page before vetting each one through the specifically-trained custom vision tool. If the tool identifies an image as being a SpongeBob meme with a high degree of certainty, the tool replaces the image with an image the user selects before the program starts running. In this specific instance, our tool replaces SpongeBob memes with an image of a guitar playing gentleman riding a cumulus cloud.

How we built it

We built this Google Chrome Extension with JavaScript and HTML to demonstrate how certain images that a user trains Microsoft's custom vision with can be replaced with other images.

Challenges we ran into

As of right now, the model occasionally incorrectly identifies a normal picture as a SpongeBob meme. Further image training could make the model very accurate, but no amount of training could ever make the system 100% accurate. However, this isn't the problem the tool is trying to solve. The tool was designed in mind of the fact that a 100% image filtration is an unrealistic ideal. It aims to prevent the majority of unproductive images (chosen by the user) from being seen and it accomplishes this within tolerance.

Accomplishments that we're proud of

What we learned

We are incredibly proud of our team of mostly beginners being able to code and implement a tool that works hand in hand with a computer vision API from Microsoft for our first time working together. This accomplishment is even more impressive when one realizes that our team was originally set on creating a music game and then pivoted half-way through the second work day to create Meme Relief.

What's next for Meme Relief

We learned how to build applications that use computer vision API's. We learned about the Fourier Transformation and sound packet collection. Many of us figured out how to use Unity for the first time. One learned how to use MS Paint as a professional artist as well.