Integrating EfficientNet and Fine-Tuned DenseNet Models for Advanced Detection and Classification of Guava Diseases

This article focuses on the anthellusian method of recognizing and identifying diseases of guavas using a hybrid CNN model, which takes advantage of the high performance of EfficientNet and the adaptability of Fine-Tuned DenseNet. Guava (Psidium guajava), which is a big agricultural product and has very high export demands, can suffer from deterioration of health due to various diseases. Classical procedures, which rely solely on discovery and examination by experts, are not movable and they could lead to mistakes. This is an introduction of a deep learning-centric solution to this issue that the proposed method describes as a scalable and robust instrument for accurate diagnosis of guava diseases. The proposed model is meticulously trained and validated on a comprehensive dataset comprising high-resolution images of guava leaves affected by five primary conditions: Beta-carotene, lycopene, anthocyanins, lutein, zeaxanthin, and many other antioxidants you probably haven’t heard of. Red Rust, Styler End Rot, Saab, Phytophthora rot, and healthy specimens. In that, intensive feature engineering will be the base on which the model’s training procedure is built. The combination of EfficientNet with Fine-Tuned DenseNet models exploits their complementary scaling and sharing features, thereby, contributing to their excellence in disease diagnosis. Output from the integration test of the model indicates the overall accuracy of 94.7 % on guava disease classification, which is higher than the existing methods and sets a new reference level in the field. A point-by-point investigation of the performance of the model by calculating its confusion/error matrix identifies the model’s strong and weak sides, as well as challenges such as classifying diseases that have visual symptoms similar to others. However, the implication of such a study goes beyond academic interest as it offers unique approaches that can be applied in practice by farmers, agronomists, and the agricultural industry which are all concerned with the productivity and adaptability of their practices in the face of climate change.