Publication

Abstract : For the diagnosis and identification of invasive pests and diseases at an earlier stage in tomato plants, an effective plant disease diagnosis system should be developed. Therefore, this article proposes an automatic tomato leaf disease identification system using a pre-trained convolutional neural network (CNN) architecture powered by crossover-based wind-driven optimization (CROWDO) algorithm. This method helps to improve the classification accuracy of tomato leaf disease when compared to the existing supervised machine learning algorithms. Here, the CROWDO approach is used to improve the accuracy of the pre-trained AlexNet architecture to reduce overfitting, training time, and excessive energy consumption. The proposed model is capable of identifying four tomato leaf diseases (Leaf Mold, Early Blight, Target Spot, and tomato Yellow Leaf Curl) and the leaf damage caused by two-spotted spider mites. The experimental analysis has been performed, and the results demonstrated that the CROWDO-optimized AlexNet architecture for tomato leaf disease classification outperforms various pre-trained such as VGG16, GoogleNet, SqueezeNet, and LeNet by the accuracy of 99.86%. The time taken for training our proposed CROWDO-optimized AlexNet architecture was 12.3654 s which is relatively less than the time taken by other pre-trained architectures such as LeNet, VGGNet, SqueezeNet, and GoogleNet in terms of seconds. The proposed CROWDO-optimized AlexNet architecture not only improves classification accuracy but also assisted the farmers in identifying the area of disease and pest infestation in real-time, as well as meeting real-time detection accuracy standards.