Test Grading App for Teachers

Our client is a teacher from North America who used to spend a lot of time grading students’ tests.

When it comes to studying, assessments are a great way to measure progress and identify problem areas. Online tests have issues with logistics and cheating, so teachers give paper tests and spend hours manually grading them. It is a repetitive, tedious, and time-consuming process that adds stress to overloaded and burned-out teachers.

As a teacher, our client used to spend lots of time grading students’ assessments. He came up with an idea of a tool that would enable teachers to eliminate manual work and save valuable time. With this idea, he approached us.

The solution is targeted at teachers in middle school and up, in both private and public education institutions worldwide. The app is especially handy for countries with large classes (e.g., India) that test heavily.

The solution uses Computer Vision and custom answer sheets to grade the assessments. In a nutshell, teachers create an answer key and students respond in these answer sheets. After that, the teacher needs to take the smartphone and point it at a hand-written test. Then the application compares the two to generate results, and instantly that test is graded. Upon the grading completion, the app sends the results directly to the students via email.

The tool recognizes not only multiple-choice tests but also all the different variations of student handwriting, including different languages and math equations. This way, the solution is tailored to any subject and is not limited to a specific type of test.

The solution consists of the following elements:

• React/Redux-based Single Page Application (SPA)
• Django JET-based Admin Panel Application
• API Application based on Django/Django REST Framework
• Computer Vision Application is based on Django/Django REST and Pillow/ Python OpenCV and consists of PDF generating, image preprocessing, MathPix integration, and MathPix results postprocessing
• Database is responsible for the datastore: users’ data, classes, answer keys, grade results, etc.

Here is how all these parties interact to perform assessments grading:

• The teacher takes a photo/photos.
• SPA sends the photos and the data about the Answer Key to the API Application.
• API Application processes the request and makes an API call to the Computer Vision Application.
• Computer Vision API receives a call and puts the task into Task Queue (Amazon SQS).
• When a task appears, Computer Vision Worker (CVW) starts processing.
• CVW makes image preprocessing (scales an image, makes it grayscale, changes the contrast, etc), and sends it to SDAPS Wrapper. The results are a CSV file with answers to multiple-choice fields and images of Textarea Answers.
• CVW combines images of Textarea Answers into a few big images and passes them to MathPix.
• MathPix recognizes text, numbers, formulas, etc on images from point 7 and returns to CVW.
• CVW combines the data and sends the result to the API Application.
• API Application grades the result of Student/Students and saves the result in the Database.
• API Application notifies the SPA that the result is ready.
• The teacher views the result.

Itexus’ cross-functional team of experts delivered the first version of the test grading app for teachers according to our client’s requirements, on time, and on budget. Now it’s available for the first real users. The client is collecting their feedback to improve and refine the app to get ready for the second round of investment.