ABSTRACT
In this chapter, we studied the feasibility of using transfer learning to identify the change process of a fresh pepper. It is very meaningful to identify the process of pepper change, especially in the degree of freshness of fruits and vegetables. As far as we know, no one judges the relationship between the quality of fruits and vegetables and time by deep learning. We tried to identify the change process of the pepper with the help of pretrained model (Googlenet). The fresh pepper collected on the same day was placed in a shaded environment, and the change of the pepper for 20 days was identified. The accuracy of 1-day, 2-day, 3-day, and 4-day, respectively, are 55.3%, 69%, 82.4%, and 96.2%. To test the generalization of the model, we also applied the model to the classification of dried pepper. In order to expand the application of transfer learning to image recognition, more studies on the changing process of fruit and vegetable varieties are needed in the future.
