DANN: A Deep Attention Neural Network for Automatic Fruit Image Classification

Classifying fruits is crucial in various fields, including agriculture, food processing, and computer vision. Typical fruit sorting often uses simple features along with basic machine-learning models. It struggles with complex fruit images. Here, we introduce a deep-learning model for fruit sorting. We propose a custom-designed CNN with an attention feature called DANN. Three standard datasets, namely, Fruits-360, FIDS30, and FRUITSGB, are used to evaluate the proposed DANN model. Testing shows that the proposed DANN model automatically classifies the fruit images with enhanced accuracies of 98.38%, 87.3% and 98% for the three above datasets, respectively. We adjust settings like batch size and epochs for training. Our CNN model with attention performs better than typical machine learning models. This research shows that adding attention to CNN improves the performance of fruit image classification. It helps in the agriculture and food industries. Our model opens paths for more research with bigger datasets and different models. This work highlights deep learning and attention for good fruit classification. It helps in farming and food quality.