Fruit Segmentation and Identification through Image Processing with K-Means and MobileNet V2

Abstract

This study presents the development of an application that integrates the K-Means Clustering algorithm and the MobileNetV2 pre-trained model to enhance image segmentation and object identification processes. Employing an experimental approach, the research incorporates Mini Batch K-Means technology to streamline image segmentation, significantly reducing computational overhead. Additional functionalities, including grayscale conversion, thresholding, and FAISS (Facebook AI Similarity Search)-based matching, are implemented to improve efficiency. The application features a user-friendly Tkinter-based GUI, enabling real-time image data upload and processing. The primary objective of this research is to optimize the accuracy and efficiency of segmentation and object identification for diverse practical applications. Experimental results demonstrate that the proposed algorithms and models achieve robust performance, establishing a foundation for the future advancement of more sophisticated technologies in this domain