Advanced Fruit Quality Assessment using Deep Learning and Transfer Learning Technique

Abstract

Ensuring the quality of fruits is essential for consumer satisfaction and food industry standards. Accurately identifying and classifying high-quality fruits is essential for maintaining good health and nutrition. The subjective and labor-intensive nature of traditional techniques of evaluating quality highlights the need for automated solutions. This study utilized transfer learning with Four pre-trained convolutional neural network (CNN) models—MobileNetV2, ResNet50, VGG16, and EfficientNetB0 —to detect the quality of five fruits: apple, banana, strawberry, orange, and mango. All models were trained and tested with a public labeled fruit images dataset, and their performance was evaluated in terms of accuracy, precision, recall, and F1-score. Our results demonstrate that ResNet50 consistently achieves the highest accuracy across all fruit types, surpassing MobileNetV2, EfficientNet and VGG16. Additionally, our models' performance is benchmarked against state-of-the-art techniques, underscoring the superior accuracy and reliability of ResNet50 in automated quality control systems within the agricultural and food sectors.