A Comprehensive Study on Classification of Desiccated Fruits Using Transfer Learning Models

Desiccated fruit and nuts are super foods that are superior to other foods. It provides a lot of vitality in the form of nutrients, vitamins, minerals, and proteins in even tiny quantities and also boosts our immunity. Desiccated fruit has been highly prized for its sweet flavor, nutritional content and prolonged shelf life. Primary dataset are used, contains 2000 images of 9 types. Five transfer learning techniques were used to address this issue as well as to reduce training time and increase classification performance. Desiccated fruit are categorised using VGG16, Xception, MobileNet, InceptionV3, and InceptionResnet. Results are compared with Convolutional Neural Network (CNN) basic model. The MobileNet model beat all other pre-trained models in terms of accuracy, achieving 99.42 % and 100 % training and validation with only 5 epochs. VGG16 also resulted in similar accuracy with a thin difference of less than 0.06 accuracy.