-
Clyra's Logo
Inspiration
A problem we see as students ourselves is that we often spend too much time studying for exams by focusing on every sub-unit covered, rather than focusing on the topics we struggled with the most. Traditional studying like this wastes valuable time before big exams like midterms and finals as well. Therefore, we wanted to create an online tool that can deliver customized insights instantly. Just as fitness trackers log all your physical activity and display the data for you to see change over time and target your weaknesses, we wanted a tool that could deliver this type of effective growth for academics. We wanted something that would use a user's past progress to determine what should be done in the present to maximize the future, just like a time machine.
What it does
The interface allows users to upload images of graded test papers, all of which will be saved under their account. An AI model will scan the paper, detecting the subject/unit of the test, the date it was taken, and the score, which it will convert to a percentage if it was written on the paper as a fraction. It will store all the uploaded test papers under their associated subject, which can be customized by the user, and generate a visual heat map highlighting weaknesses. This type of analysis will help students focus their time on what topics they are struggling with.
How we built it
We started with the front-end interface, using mainly HTML and CSS to form a landing page, a dashboard, a scanner tab to upload the photos, and a login/registration form. We kept a coherent "futuristic" theme with varying shades of purple to add to the concept of an academic time machine. We used some JavaScript code to add some functionality in certain areas, although it wasn't a main part of the code. After this, we created the back-end code using Python to store data created in the registration page and save it in the login page, allowing people to create and log in to accounts easily. We used an AI model through OCR, which we programmed to detect subjects, dates, and scores on exams, training and testing it with some of our past year papers. We then used Python libraries to make the heat maps on the Dashboard tab, which would take the data from the uploads and turn it into a visual interpretation. We wrote and compiled all these files on Visual Studio Code.
Challenges we ran into
The OCR proved a very complex challenge as, originally, it did very poorly at reading scanned documents. Therefore, we had to spend a lot of time (almost all of Saturday!) reprogramming and testing the OCR to allow it to properly read the information on the page. It took a lot of time and effort, we were able to make it almost exact at reading documents with 90 %+ accuracy. Designing an intuitive user interface was also difficult, with the back-end code giving us a lot of trouble. It was difficult linking Python code to HTML code, to the point where we even tried switching to TKinter to make the integration process easier. However, we knew that option wouldn't give us the flexibility and intuitiveness that HTML and CSS provided, which is why we decided to make it work, which eventually, it did. Lastly, it was difficult staying within what was feasible for the time we were given, as we were very excited about our idea and thought of many more features to add, but we knew we had to stick with what we could do within the scope of our abilities and the time frame.
What's next for Clyra
In the future, we hope to better train our AI model to detect different types of handwriting, as sometimes it isn't able to read some forms of handwriting. With more time, we could not only improve our AI model, but also program our software to use the weaknesses to form study plans. This would bring less work to the user and make their studying much more efficient. We hope to formally publish our program one day, but we want to make sure it remains completely free of charge to anyone who wants to use it. As a student, sometimes online websites make life easier, but it becomes increasingly difficult when everything is blocked by a paywall. We wish to prevent that problem for users like us, while still delivering a quality product that we hope will make their lives easier.
Log in or sign up for Devpost to join the conversation.