Project DeepScan started out in the belief that every single person on Earth should be provided with the equal educational opportunities. However, different levels of resources provided for instructors in different schools limit the learning/teaching opportunities in schools with smaller budgets.
What it does
Project DeepScan seeks to narrow a small part of this gap of educational opportunities between schools by providing automation of grading processes with machine learning. The project hopes that the instructors will have more time with their students by taking the burden of manual grading off of their back.
How we built it
Project DeepScan is implemented with the neural net and deep learning algorithms.
Challenges we ran into
Setting up the machine learning models usually takes more than a few hours, and we spent good amount of time during the hackathon just doing that.
Accomplishments that we're proud of
The accuracy of the machine learning algorithm in recognizing handwritten letters are ridiculously high. With little more time, we should be able to achieve the accuracy close to 99%.
In the implementation, Project DeepScan does not have any complicated piece. The project brings several very simple ideas from different technologies together to create something extra-ordinary.
What's next for DeepScan
Improving the accuracy of handwritten letters are very-near-future goal. However, the scope of the problems that DeepScan can handle seem to be limited to multiple choice questions for now. In the long run, we highly intend to provide automated grading even for questions with short answers.