This study presents the design and implementation of an automated system for sorting and measuring kidney tomatoes using a YOLOv8 model with a size estimation algorithm. The proposed system integrates computer vision and deep learning with a physical sorting mechanism to categorize tomatoes into three classes: green, red, and damaged, while also determining their size. The classification model was trained on a dataset of 2,145 images of tomatoes taken from different sources and lighting conditions to enhance performance during training. The implemented prototype consists of a conveyor belt equipped with sensors and a high-resolution camera to capture and analyse tomato characteristics in real-time. A servo-driven sorting mechanism then directs the classified tomatoes into their respective bins. Experimental validation and testing show that the model achieves a classification accuracy of 99.6% and a size estimation accuracy of 97.1%, aiding in a reliable and efficient post-harvest sorting process. The proposed system not only reduces the probability of human error but also improves the precision of tomato classification. Future developments will focus on refining and adapting existing AI methodologies to improve their effectiveness in agricultural environments. This includes enhancing model robustness, improving classification accuracy under real-world conditions, and tailoring AI tools to better meet the demands of industrial tomato sorting
