writerDecipher

Inspiration

The inspiration for this project came from a scenario that I'm sure we're all familiar with: a classmate didn't write their name on an assignment and the teacher has to beat around the bush trying to figure out who the author was. What if this embarrassing and time-wasting practice could be done away with?

What it does

The end user simply takes two pictures on the app and the app does the rest! The connected python script will parse through the recognized characters and compare each like character. If the net comparison results in high correlation between the two samples, the app will report to the user that the samples likely came from the same author, as well as the calculated probability that the authors are the same.

How we built it

In terms of the script, we started with the an implementation of the Google Cloud Vision API to determine the text that was written. Next, we used recursion to find and isolate each individual letter. Finally, for the comparison, we sorted all pixels into RGB categories and employed a pixel by pixel comparison. For the app side of things, we used android studio to code a simple app that would facilitate the image-taking and processing. Using flask, we were able to connect the script to the app and send data from user to server.

What we learned

Image recognition and comparison is really hard to implement. Seemingly trivial tasks can be made a whole lot more difficult when dealing with image data.

What's next for this project

The solution is, as of now, relatively fragile. Going forward, a database system would need to be implemented to allow the storage of long-term handwriting data. Additionally, the image detection and comparison systems would need work so that they could handle a wider variety of text data.