OCR handwritten text for Educational Institutions
We created a machine learning models able to recognize handwritten text of pupils' names and their marks, in order to be entered from a paper register to a web based application.
Problem
All schools in Moldova must have a paper register with marks. It takes a couple of days for the teacher to calculate the average marks at the end of the semester. Also in the holiday, a special team is created in order to identify the mistakes teachers made.
Solution
We trained a machine learning model able to classify marks in 13 classes, and able to recognize handwritten names of the pupils from the marks table. For the classification of the mark, we used state of the art model like resnet. For the name recognition, we generated thousands of examples and annotated a couple of hundreds, and got an accuracy of 87% per name.
