Abstract:
The "Deep Learning Techniques for Automated Detection of Diseases in Guava Fruits
and Leaves" research project aims to revolutionize guava farming by creating an
automated system for early disease identification. By harnessing cutting-edge deep
learning techniques, we strive to significantly improve agricultural production by
accurately diagnosing plant illnesses. Our extensive collection of tagged guava plant
images will lead to the training of advanced deep learning models, ultimately
benefitting guava farmers worldwide. Our evaluation of various models, including
ResNet50, VGG16, Inception V3, and DenseNet121, revealed that the ResNet50
model outperformed the rest with an impressive test accuracy of 92%. Its unparalleled
accuracy and robustness make it an ideal choice for practical implementation in guava
farming, promising substantial advancements in disease detection. Capturing and
analyzing data, creating models, deploying solutions, and rigorous testing are pivotal
project phases essential for success. Our meticulously planned project management
includes a well-structured calendar, clear work allocations, and key milestones to
ensure efficient and timely execution. A comprehensive financial analysis forecasts
the project's total cost to be between 75,900 and 90,900 taka over a six-month period,
encompassing field trips, personnel, software, hardware, cloud services, and other
related expenses. This study aims to deliver a reliable solution for guava growers,
enabling timely and accurate identification of plant diseases, ultimately leading to
healthier crops and increased yields
