Coffee Green Bean Quality Classification using GoogLeNet
Masud Effendi1*, Januar Wahyu Bagas Firmansyah1, Usman Effendi1, Imam Santoso1, Retno Astuti1, Wayan Firdaus Mahmudy2

1 Department of Agroindustrial Technology, Faculty of Agricultural Technology, Universitas Brawijaya, Malang, Indonesia
2 Faculty of Computer Science, Universitas Brawijaya, Malang, Indonesia
*mas.ud[at]ub.ac.id

Abstract

Coffee is a commodity that plays an essential role in the Indonesian economy. Creating a quality identification model can help farmers get the best quality from every coffee they make. This study presents the development and evaluation of a deep learning model using GoogLeNet to identify coffee quality. The model aims to classify coffee beans based on visual quality parameters. A dataset of 2600 coffee bean images was collected and preprocessed. The GoogLeNet architecture, leveraging transfer learning, was trained and fine-tuned. The architecture was tested by varying the batch size, learning rate, and optimizer to select the best accuracy. The best model results were obtained on the GoogLeNet architecture with the Adam optimizer using a batch size of 64 and a learning rate 0.001. The accuracy obtained was 85.4%. This work contributes to improving automated coffee quality assessment for the coffee industry.

Keywords: Deep Learning- Coffee Quality- GoogLeNet- Image Classification

Topic: Smart technology for sustainable agro-industry