Abstract

ABSTRACT – Bananas are plants from the Southeast Asian region and belong to the genus Musa and the family Musaceae. Grown in tropical and subtropical regions, bananas are one of the agricultural commodities with the largest production compared to other fruits. Indonesia is one of the countries that produce the largest bananas in the world. The yields are then sorted based on the level of ripeness by looking at the color change of the banana skin. However, the process of sorting bananas requires a lot of time and effort due to the large production of bananas. In addition, differences in the assessment of each individual on changes in the color of banana peels result in an unstable or consistent sorting of bananas. Therefore, this study intends to create a ripeness classification system for bananas based on changes in skin color with the aim that the sorting process can be carried out efficiently and accurately. The color variants used range from dominant green for unripe bananas, dominant yellow for ripe bananas and blackish brown spots for overripe bananas. The method used is a Convolutional Neural Network with a self-designed architecture. The results showed that the accuracy reached 88% with a learning rate setting of 0.001 and a maximum epoch limit of 15.


Keywords – Classification; Banana Fruit; Convolutional Neural Network; Deep Learning; Computer Vision