APPLY A PRE-TRAINED CONVOLUTIONAL NEURAL NETWORK MODEL TO OPTIMIZE DISEASE CLASSIFICATION IN COCOA

The research presents the concept of using pre-trained convolutional neural networks to develop models. This research has the objective to create a model for disease classification in cocoa from images using a pre-trained convolutional neural network technique. and test the performance of the model. The first round of training did not adjust the data, and the second round had data adjusted by Brightness, Zoom and using a combination of Brightness and Zoom. The results of the experiment found that the Xception model passed. Data adjustment using a combination of Brightness and Zoom had the best performance in classifying diseased and non-diseased cocoa fruit images with the highest accuracy value of 0.91, followed by the InceptionV3 model with an accuracy value of 0.84.