What it does

Have you ever wanted to take your notes from a class or meeting and put them online without the hassle of typing it all up? This app allows you to take a picture of your notes and convert them directly to text through a simple online website. You can then take the transcribed text and use it in any application you'd like.

How we built it

We used an open-source machine learning model that was trained to detect sections of text, and then feeds into a second layer model that detects sentences, which then flows onto a third model to detect the words and predict what letters are being used. The predictions are cross-referenced with a dictionary to increase accuracy and improve sentence structure. The pictures are uploaded to our Django web-application, run through OpenCV for rectification, straightening, black-and-white conversion, and fed through the neural network. This neural network generates a text file, which the web-application delivers to the user.

Challenges we ran into

One challenge we ran into was that the .tech domain did not work with Google Cloud Platform. We also needed to tweak the handwriting recognition algorithm extensively in order to have it run accurately.

Accomplishments that we're proud of

We were able to build a Django web application even though none of our team members had any experience with web development (frontend or backend).

What we learned

We learned how to use Django, HTML, CSS, and enhanced our OpenCV skills.

What's next for Diginote

In the future, we see Diginote as an appication that every student uses to digitize their notes. We plan to support digitizing tables, flowcharts, and more in the future, and hope to improve the accuracy of the handwriting recognition algorithm.

Credit to https://github.com/ThomasDelteil/HandwrittenTextRecognition_MXNet for the trained model weights We intended to use, and obtained the domain "diginote.tech," but the .tech verification was not functional.

Share this project: