Fruit Classification with Deep Transfer Learning using Image Processing

This research aims to assess the effectiveness of fruit classification through a combination of transfer learning and image processing techniques. A dataset comprising 216 diverse fruit images, such as durian, graviola, pupunha, acai, guarana, and cupuacu, was collected and split into 180 images for training and 36 for testing. In this study, we introduce a deep learning-based approach for fruit classification, leveraging three Convolutional Neural Network(CNN) models: ResNet-50, VGG-19, and MobileNetV2. These models were employed to evaluate and compare the performance of fruit classification using transfer learning and image processing methods. Notably, MobileNetV2 achieved the highest verification accuracy of 97%, outperforming ResNet-50 with a test accuracy of 87% and VGG-19 with a test accuracy of 17%. This approach offers improved efficiency, simplicity, and accuracy in fruit sorting tasks.