Application of CNN in the Classification of Chili Varieties for Agricultural Efficiency

This research focuses on the problem of classifying chili harvests which is still done manually by farmers. This manual classification process will of course take a long time, require a lot of energy and will feel tedious. This research aims to develop a classification system for chili types using the Convolutional Neural Network (CNN) method. By utilizing CNN technology, it is hoped that the chili grouping process can be carried out automatically with a high level of accuracy, thereby increasing work efficiency and reducing errors in chili grouping. The data used in this research is primary data with a total of 500 images of chilies divided into 4 classes. These images were taken using a Samsung A7 smartphone camera under consistent conditions: all photos were captured during daylight hours with the same camera angle. The training and testing results of the CNN model in classifying types of chili showed an accuracy of 99.5% in the training stage and reached an accuracy of 94% in the testing stage. Based on these results, it shows that the application of the CNN method in classifying chili types can work very well and effectively.