Guava Leaf Disease Detection in Bangladesh Using Deep Learning

Abstract

Guava is a tropical fruit that is both nutritious and economically valuable, and it is widely grown in tropical and subtropical regions. However, its production is often affected by serious diseases such as Red Rust and Algal Spotting, which can greatly reduce both yield and fruit quality. Early accurate detection of diseases is very important for timely crop management. However the traditional methods of detecting diseases such as resort to experts for manual inspection are not only time consuming, labor consuming, but also not applicable for the large-scale production. Farmer detect disease using their eye sight. It is not very good practical solution. Farmer Recent developments in deep learning have introduced new prospects for agriculture with efficient and automated disease detection systems. This work employs deep learning methods for the improvement of yield and quality of guava through the development of an automated technique that focuses on the detection of diseases in guava leaves and differentiation of healthy leaves. In this study we gather a total of 4,887 labell ed images of guava leaves. Five state-of-the-art deep learning models (including DenseNet201, InceptionResNetV2, VGG16, VGG19, and MobileNetV2) were trained and tested on this dataset. These models were selected because they are widely used in image classification and have shown good results on agricultural data. We tested them by measuring accuracy, precision, recall, and F1 score. Among the models, DenseNet201 performed the best with an accuracy of 96.25%, followed by MobileNetV2 with 95.64%. The other models, such as VGG19 and VGG16, also gave strong results. These findings show that deep learning can successfully identify guava leaf diseases and separate them into different categories. This helps farmers detect problems early, control diseases more effectively, and improve crop yields. The goal of this research is to build an automatic system that can assist farmers with disease detection and crop management. In the future, this work can be combined with IoT devices to make detection faster and more practical. Future improvements will also include adding more diverse data to cover additional cases and building a real-time detection system that can support larger farms. Overall, this research shows how AI can make farming smarter, more efficient, and more sustainable.