Next project: Leaf Species Classification
When it comes to grading assignments all the tasks were manual at University of Dayton. It was a tough job to grade assignments of 20-30 students every week, and that it would be great if there was a system that could automatically grade the assignments. Before even the development began the idea had some problems that had to be dealt with:
The idea was to develop a system that would be beneficial to all the TA's of Computer Science department. Since the department is huge the problem was to figure out how all the TA's of different courses should be given access to the system while students having to log in into only one account for all registered courses.
Assignments have to be manually checked for plagiarism even while the grading is done automatically. Each and every file that is uploaded by the students has to saved for the TA's to later verify that the code hasn't been copied from other students.
Students would like to use the system but they might not want to provide any personal information. The problem arises as to how the students can use the system without having to provide any personal information. The system has to take care of all the information and code that is provided by the students.
After detailed planning and identifying individual component within the system and the ways to keep user data secure, it was then possible to design each section of the Auto Grader. The sections that bring Auto Grader to reality are:
The home dashboard was designed to be simple and do only one job, that is to display all the information of the courses that the students are registered to. The courses on the home dashboard are dynamically fetched from the database, so in the future when students are added to any new courses, those courses will automatically reflect on the home dashboard.
When the students navigate to the course, they are redirected to the course dashboard. The course dashboard is also dynamic and displays detailed information of all the assignments that are open to students, assignments can only be accessed until the submissions is valid. The students are provided with statistics about their courses; such as the number of submissions in each course, the total grade of all the courses is also provided so that students can focus more on the courses that they are not doing good at.
The end point for the students is the assignment dashboard where they can upload a zip file that contains all the project files necessary to run their assignment. The assignment dashboard allows students only to the upload files for "n" number of times, where "n" is defined by the TA for every assignment. The TA defines the number of times an assignment can be graded by the level of their education (Undergraduate or Graduate level) and by difficulty level of an assignment.
Once the files are uploaded, then the zip file is unzipped to extract all the project files. There are multiple checkpoints when it comes to generating a report; firstly the number of files uploaded and their naming convention are checked, secondly when the files run auto grader checks for any infinite loops and terminate the program if any, and finally a detailed report is generated and saved locally on the server and mailed to the students using NodeMailer. Report provides students with their final scores and suggestions to improve their assignment.
Auto grader was complex to design and it took weeks if not months to design and develop. It was developed using technologies that seem simple and uncomplicated, but auto grader as simple as it seems to understand was very difficult to implement with the aim to solve three major problems. Piloting this project was the best decision I made in 2020 because I believe that I could learn how to bring ideas to reality. Auto grader was a simple problem statement which seemed easy to solve but it's only when you really plan the project one comes to know the complexity of the problem.