Deep learning approches to banana ripeness detection

Thanyacharoen S., Jirapattarasakul S., Joijinda N., Leone G. R., Akkaralaertsest T., Yingthawornsuk T.

Deep Learning, Image Processing, Banana Ripeness Classification, VGG16, VGG19, InceptionV3, Xception, MobileNetV2 

This study presents a comparative analysis of five deep learning models VGG16, VGG19, Xception, InceptionV3, and MobileNetV2 or banana ripeness classification. A dataset used in this study consists of four levels of ripeness: overripe, ripe, rotten, and unripe. Each deep learning model was fine-tuned using pre-trained weights from ImageNet to adapt them for banana ripeness classification each model was fine-tuned with pre-trained weights from ImageNet. The evaluation was conducted using 5-Fold Cross-Validation to ensure the robustness of the results. As results shown, the VGG16 achieved the highest accurate performance of 93.7%, surpassing other models in all metrics. MobileNetV2 and Xception followed closely, demonstrating competitive results, while InceptionV3 had the lowest accuracy. Results indicate that VGG16 is best suited for banana ripeness. This research highlights the potential of deep learning in automating fruit ripeness detection and provides valuable insights for agricultural applications. Future work may explore the integration of additional datasets and real-time deployment for broader use cases.

Publisher: IEEE

Metrics