Deep Learning Models Performance for Classifying Dried Chili Based on Digital Image Analysis

Abstract

Artificial Intelligence and agriculture have been combined to create significant advancements in smart agricultural analysis that have improved both output and quality. This approach has completely changed conventional farming operations by utilizing image processing technologies. To assess the dryness levels of red chili peppers—a crucial component of crop quality and market value—the study set out to compare the efficacy of several CNN architectures. A dataset with 600 training images and 150 testing images spread over three classes was used to evaluate four CNN models (MobileNetV2, DenseNet121, InceptionV3, NASNetMobile). With a validation accuracy of 99%, DenseNet121 outperformed MobileNetV2 (which had a validation accuracy of 97%). The findings demonstrate how deep learning models can improve sorting procedures for agriculture by increasing accuracy and productivity. A scalable, impartial, and economical way to uphold crop standards and promote industry sustainability is by incorporating CNNs into the classification of agricultural products. The results of this study represent a breakthrough in the application of deep learning to agriculture, opening the door to automated systems that guarantee constant product quality. By optimizing yield and quality through image processing technology, the findings highlight the revolutionary influence of AI in smart agriculture. To increase production and improve competitiveness in the market, future research efforts may focus on developing automated sorting systems and further enhancing CNN models for agricultural applications. The research adds to the increasing corpus of work using AI in agriculture to enhance crop management and quality control.