This paper presents a profound learning-based approach to classifying nearby Bangladeshi bananas and surveying their readiness utilising the NasNetMobile demonstration. Conventional strategies of sorting bananas are manual and regularly wrong, making robotisation vital for moving forward effectiveness in agribusiness. We built a dataset of 2,082 pictures, capturing bananas in several lighting conditions and readiness stages, and connected information increase to upgrade demonstrate execution. After testing a few profound learning models, counting Xception, MobileNetV3Small, InceptionV3, and InceptionResNetV2, NasNetMobile risen as the most excellent entertainer, accomplishing 97.12curacy. Its lightweight plan and solid highlight extraction make it appropriate for real-world applications, counting mobile-based classification instruments for agriculturists. Whereas the show performs well, challenges like natural varieties and real-time arrangement stay. Future advancements incorporate extending the dataset, optimising for portable gadgets, and joining half-breed AI strategies for way better precision. This research takes a step forward in exactness agribusiness, advertising an adaptable, AI-driven arrangement to assist agriculturists and supply chain supervisors in making educated choices, almost gathering and offering bananas.