Fruit and vegetables classification system using image saliency and convolutional neural network

Fruit and vegetables classification and recognition are still challenging in daily production and life. In this paper, we propose an efficient fruit and vegetables classification system using image saliency to draw the object regions and convolutional neural network (CNN) model to extract image features and implement classification. Image saliency is utilized to select main saliency regions according to saliency map. A VGG model is chosen to train for fruit and vegetables classification. Another contribution in this paper is that we establish a fruit and vegetables images database spanning 26 categories, which covers the major types in real life. Experiments are conducted on our own database, and the results show that our classification system achieves an excellent accuracy rate of 95.6%.