Maulana, Reihan (2023) Pemanfaatan Augmentasi Data dalam Klasifikasi Handwritten Digit Images Menggunakan Algoritma Cnns Berbasis Android. Other thesis, Universitas Islam Riau.

Text 193510037.pdf - Submitted Version Restricted to Registered users only Download (9MB) | Request a copy

Abstract

Convolutional Neural Networks (CNN) have demonstrated their effectiveness in handwritten digit classification, achieving excellent accuracy on the MNIST dataset. However, as is often the case in deep learning, the issue of overfitting arises when the training data lacks representation of samples from the data distribution that will be tested. Overfitting models tend to perform well on the training data but poorly on the testing data. Data augmentation techniques can be employed to increase the sample size by transforming and modifying the original data using techniques such as rotation, translation, and zoom. Testing was conducted by comparing the performance of models before and after applying data augmentation in the Handwritten Digit Classification Application, considering positions such as top, bottom, right, left, and center. The results of the testing demonstrated that data augmentation techniques effectively address overfitting and improve the performance of the handwritten digit classification model.

Actions (login required)

View Item