Identifikasi Kualitas Visual Rempah Ekspor Indonesia Menggunakan Deep Learning Berbasis CNN

Abstract

Indonesia is the world's leading producer of spices, but it still faces challenges in manual visual quality assessment, which is inconsistent. This study aims to develop a spice quality classification system using a Deep Learning approach based on Convolutional Neural Networks (CNN). Data was collected through digital images of five types of spices (cloves, cardamom, cinnamon, pepper, and nutmeg) classified into two categories: good and bad. The dataset was then processed and used to Train the CNN model using Tensorflow. The model architecture consists of several convolution, pooling, and dense layers, and is integrated into a web-based prototype application using Streamlit. Evaluation results show that the model achieves high Accuracy of 98.86% (Training), 98.45% (Validation), and 98.45% (Testing). The prototype application can provide automatic Predictions of spice quality through a simple and responsive interface. The results of this study indicate that CNN is effective in identifying the visual quality of spices and can serve as an objective, efficient technological solution that supports the enhancement of Indonesia's spice export competitiveness.