The Role Of Gray Level Co-Occurrence Matrix In Convolutional Neural Network Transfer Learning For Coffe Bean Classification

Main Article Content

I. Kadek Dwi Nuryana, Herlina Syafhita Maharani, Zainal Ikhwan Muhammad

Abstract

This research explores the impact of the Gray Level Co-Occurrence Matrix on enhancing Convolutional Neural Networks for categorizing coffee beans, focusing on a case study from Permadi Pandansari. The data used in this study is training data for 2 types of coffee beans of 400 images, while test data for 2 types of coffee beans is 100 images, so the total is 500 images of coffee beans. The extraction process used is the extraction of texture and color obtained from the extraction of the Grey Level Co-Occurrence Matrix. Followed by the deep learning method used for grouping is Convolutional Neural Network using VGG-16 transfer learning. To maximize the results, this study also applies ADAM optimization and also ReLU and Softmax activation. The results of the feature extraction test are determined by the values of accuracy, precision, recall, F1-Score and also Cross Validation. This research investigates the influence of the Gray Level Co-Occurrence Matrix on enhancing Convolutional Neural Network performance in coffee bean classification, centered on a case study from Permadi Pandansari. The dataset comprises 500 coffee bean images, with 400 images for training and 100 for testing, representing two types of coffee beans. Texture and color features were extracted using the Gray Level Co-Occurrence Matrix, followed by classification using CNN with VGG-16 transfer learning. To optimize performance, ADAM optimization, along with ReLU and Softmax activation functions, was applied. The effectiveness of the feature extraction was evaluated through accuracy, precision, recall, F1-score, and cross-validation metrics.

Article Details

Section
Articles