Table 3. Research objectives to research contribution mapping.

Research objectives	Contributions
To develop fruit classification models using pre-trained models	Three pre-trained models namely VGG16, ResNet50 and MobileNetV2 are utilized for the development of fruit classification models. See Figs. 5 and 6
To explore transfer learning mechanism in the development of fruit classification models	Two transfer learning mechanisms (1) frozen layers and (2) fine-tuned layers are employed in the development of classification models. Refer to Section Phase I - Development of Classification Models for further details
To compare performance of fruit classification models using different metrics	The classification models employing pre-trained models are evaluated using accuracy, precision, recall and f1 score. See Tables 4 and 5
To interpret fruit classification models using explainable AI tool	Explainable AI tool LIME is used to interpret the results of pre-trained models based classification models. See Table 7 and Fig. 8