Handwritten Image Classification Project

Date: 2025-10-21

Prepared by: Çiçek Akkaya

Grade of the Final Exam Project: 83/100

This is Çiçek Akkaya's final exam project of Deep Learning course in Data Science and Artificial Intelligence Master's Degree Program of European Higher Education Institute, Malta.

Please browse the Cicek_Akkaya_DL_Final_v1.pdf file for the report of the project.

Please browse the DL_Final_Cicek_Akkaya.ipynb file for the python codes used.

This project ensures to recognize correctly the 10 classes of digits which are from 0 to 9, in the MNIST Handwritten Digit dataset which includes in tens of thousands of handwritten images. This is an Image Classification project which is a kind of Computer Vision variety. In the project, used CNN architecture of Deep Learning.

Training set and testing set were separete. They used separately for training and evaluating the model.

The rate of per number, for the training set and the testing set, were changing. However they were not meaningful differances. Therefore, we can say the rate of the training set approximatelly 85% and testing set is 15%.

In this project, used CNN architecture which is a Deep Learning Architecture to process and make predictions from images.

Model performance: The accuracy of the model is 99.52% which means the model has high performance for predicting class of the digits.