Modified Grouped Convolution-Based EfficientNet Deep Learning Architecture for Apple Disease Detection

Early detection of apple diseases is vital to prevent widespread crop damage and financial losses. Deep learning models have shown promise in automating the detection of diseases in plants. The proposed research seeks to address several critical challenges in the domain of apple disease detection. First, it recognizes the need for models capable of detecting diseases accurately and efficiently. Second, it acknowledges that the traditional depthwise convolutions may not fully capture the underlying patterns in apple leaf images, leading to suboptimal performance. To enhance disease detection accuracy in apple leaves, the EfficientNet-B0 architecture is modified through the replacement of depthwise convolutions with grouped convolutions. In rigorous experimentation, our proposed model exhibited better performance, achieving a training accuracy of 100% and a testing accuracy of 100%. These accuracy scores underscore the model's proficiency in both learning from training data and generalizing to the unseen data.