Comparative Evaluation of Activation Functions for Bacterial Spot Disease Classification in Bell Pepper Plant Leaves

In this research, we investigated the performance of Leaky ReLu, Parametric ReLu (PReLU), Scaled Exponential Linear Unit (SELU), and Swish, to classify bacterial spot illnesses that are seen in the leaves of bell pepper plants. Using Visual Geometric Group-19 (VGG-19)—a deep convolutional neural network architecture, we conducted a series of exhaustive experiments to determine the influence of various activation functions on the precision of disease diagnosis. The results of our research indicate that, among the activation functions investigated, Leaky ReLu regularly beats the others in terms of classification accuracy when diagnosing bacterial spot illnesses. Based on the findings, it is clear that the selection of the activation function is an essential factor in improving the overall performance of deep convolution neural networks (DCNN) when applied to this particular application. This research contributes to the optimization of deep-learning models for plant disease detection. Categorizing diseases that affect bell pepper leaves brought to light the necessity of choosing a suitable activation function, with Leaky ReLu as the most effective option for the VGG-19 architecture.