Fruit Classification and Ripeness Estimation using Deep Learning Models

Fruits play an important function in our diet by providing essential fibers, vitamins, and minerals. Accurate detection of fruit ripeness can enhance quality control and reduce food waste. This study focuses on using CNNs to grade fruits based on their ripeness levels into three groups: unripe, ripe, and overripe. The dataset comprises images of grapes, papaya, pomegranate, and strawberry, sourced from the Fruits 360 dataset on Kaggle, RoboFlow, and other websites. Various CNN models were employed, including DenseNet201, MobileNetV2, InceptionV3, and VGG16. DenseNet201 obtained the best accuracy of 92%, followed by MobileNetV2 at 88%, InceptionV3 at 74%, and VGG16 at 68%. The proposed work demonstrates the potential of CNNs in automating the practice of detecting fruit ripeness, contributing to advancements in agricultural technology and improving supply chain management.