Innovative Solutions for Bean Leaf Disease Detection Using Deep Learning

Rapid developments in the application of Deep Learning for plant disease detection have made a significant contribution to the agricultural sector. This research is focused on disease detection on bean leaves by classifying 1295 image data into 3 classes. We propose a new Convolutional Neural Network (CNN) architecture based on DenseNet121 with modifications to its head. These modifications adding a Convolutional layer with filter 64, kernel size 3 × 3, and ReLu activation. Next, a Convolutional process was carried out with filter 32, kernel size 3 × 3 and ReLu activation, followed by the use of Average Pooling as a whole. In this research we also carried out a comparative evaluation between the proposed model and other transfer learning architectures. The research results show that the proposed model has the best accuracy reaching 96.90%, with an F1 Scor, Precision, Recall of 97%.