학술논문
Banana ripeness classification using transfer learning techniques
Document Type
Conference
Source
2022 International Conference on Computing, Communication, Security and Intelligent Systems (IC3SIS) Computing, Communication, Security and Intelligent Systems (IC3SIS), 2022 International Conference on. :1-6 Jun, 2022
Subject
Language
Abstract
Bananas are high in both water and fiber, which help to maintain regularity and digestive health. Consumers and fruit industry firms are concerned about the quality of fresh bananas, which is quickly becoming a vital need for standardizing the quality of commercial bananas. Farmers will benefit from the capacity to determine the freshness of banana fruit to increase the cropping phase and prevent the harvesting of either under-matured or over-matured bananas. When it comes to classifying photographs into their appropriate categories, machine learning plays a critical role. The study applies five different Transfer Learning models to classify banana images into three categories namely “Unripe”, “Ripe” and “Overripe”. The dataset used in this distribution contains 300 images which are further augmented to a total of 2369 images using different Image Augmentation techniques. The models used in the study include VGG-16, VGG-19, InceptionV3, Xception, and DenseNet201.The results show that VGG-16 achieves the highest accuracy in classifying the bananas as compared to the other models used in the study. The best accuracy achieved is 98.73%.