Implementation of Data Augmentation Using Convolutional Neural Network for Batik Classification

Abstract

CNN has the ability to detect and recognize objects in an image, also can outperform traditional methods in computer vision and pattern recognition tasks. CNN has been implemented in many kinds of work, one of which is to do batik classification. We implemented eight kinds of data augmentation on the batik dataset using the VGG16 pre-trained model with fine-tuning methods. The dataset consists of 500 batik images divided into five classes, namely Ceplok, Kawung, Lereng, Nitik, and Parang. Batik classification by selecting data augmentation was successful in increasing the accuracy of 3.13% from 95.83% (without data augmentation) to 98.96% (by selecting data augmentation).

Topics & Concepts