Fig Fruit Image Segmentation using Threshold, K-means Clustering, and Sharp U-Net Techniques

In this study, image segmentation on Ficus Carica (fig) was developed. Fig fruit image segmentation separates fruit objects by removing the background in the image, including shadow images, and extracting the fruit shape. The developed methods for fig image segmentation were evaluated to identify how well the methods work and were compared to find the best method for fig image segmentation. As a reference, ground truth was made using software called Procreate for comparative purposes. There were three methods used in this paper that include Threshold, K-means clustering, and Sharp U-Net. The platform used for this development is Google Colab. Based on the results obtained, the Sharp U-Net demonstrates the highest value of accuracy as compared to the Threshold and K-means Clustering techniques. Therefore, the most effective and efficient method to use on fig fruit image segmentation is the Sharp U-Net method.