Identifikasi Varietas Jagung Menggunakan Metode Convolutional Neural Network

Abstract

Corn is one of Indonesia's primary commodities, but its productivity is significantly affected by environmental and climatic conditions. Crop yields often fluctuate due to climate change, particularly the El Niño phenomenon, which can lead to drought and delays in planting seasons. As a result, selecting superior drought-resistant seeds is essential, especially in dryland areas like East Nusa Tenggara. Traditional methods for identifying seeds, such as manual morphological measurements, are considered inefficient in terms of time and labor. This study aims to develop a corn seed variety classification system using a Convolutional Neural Network that focuses on five varieties: NK 212, NK 7328 Sumo, P 21, Pertiwi 2, and Pertiwi 6. We designed and tested three CNN architectures with varying complexities. The most effective model, which includes three convolutional layers and two fully connected layers with dropout, achieved an optimal performance of 89,20% accuracy on the test data.