Fruit and Vegetable Classification Using MobileNet V2 Transfer Learning Model

Abstract

Fruits and vegetables possess considerable economic worth as genetic resources in agriculture. The motivation for this research comes from the necessity to achieve accurate and efficient classification of fruit and vegetable images. This study proposed a MobileNetV2 model architecture for categorizing fruits and vegetables into 36 groups. The aim is to explore the capabilities of this model, which has already demonstrated promising outcomes in earlier studies. The work utilizes a dataset including 3,825 photographs, and the model undergoes extensive training for ten epochs, employing a batch size of 32. The training process is optimized using the default learning rate parameters and the Adam optimizer. The investigation findings demonstrate a notably favorable outcome, indicating that the proposed MobileNetV2 model achieved an impressive accuracy rate of 96%. These discoveries will have a significant influence on the food and agriculture sectors. Developing a highly accurate and efficient automated system for classifying fruits and vegetables holds significant promise in revolutionizing various aspects of agricultural monitoring, quality assurance, inventory management, and retail automation. This study contributes to developing resource-efficient and sustainable practices in the food and agriculture sectors by enhancing categorization accuracy. Ultimately, both consumers and the industry are poised to benefit from the potential of reducing waste, improving product quality, and increasing the availability of high-quality fruits and vegetables.

Topics & Concepts