<![CDATA[Chili peppers (Capsicum annuum L.) are an important horticultural commodity in Indonesia with high economic value, but they are susceptible to leaf diseases such as leaf spots, curled leaves, yellowing leaves, and whitefly pests. This study aims to classify chili leaf diseases using a MobileNetV2-based Convolutional Neural Network (CNN) architecture utilizing the Depthwise Separable Convolution mechanism for filter decomposition and model complexity reduction. Based on previous studies, MobileNetV2 has been proven to maintain a highly competitive level of accuracy. The dataset used consisted of 6000 images from five categories: healthy, leaf spot, leaf curl, yellowish, and whitefly, which were taken from open sources and equalized in number for each class. The data was divided into training, validation, and testing sets with a ratio of 80:10:10. The training process used depthwise separable convolution, dropout, and Adam and SGD optimization techniques to prevent overfitting. Model evaluation was carried out through 12 scenarios with variations in batch size, dense layer, optimizer, and epoch. The results show the highest accuracy of 98.40% in the scenario with a batch size configuration of 32, a dense layer of 128, a learning rate of 0.001, an Adam optimizer, and 20 epochs. Most scenarios achieved an accuracy above 96%, proving that MobileNetV2 is effective for classifying chili leaf diseases. The contribution of this study is the identification of an optimal and efficient MobileNetV2 parameter configuration for chili leaf disease classification.]]>
