AI Solutions for Grilled Eggplants Sorting: A Comparative Analysis of Image-Based Techniques

In the rapidly evolving food industry, the sorting process plays a pivotal role in maintaining the quality and efficiency of production lines. Traditional methods, reliant on human operators for the selection of food items based on external characteristics, are increasingly being supplemented by advanced technological solutions. This study presents a comprehensive comparison of various Artificial Intelligence (AI) techniques applied to image analysis for food sorting, with a focus on the detection and classification of non-conforming eggplant slices in an industrial setting. Utilizing a dataset of 1015 images captured from an eggplant production line, we explore the application of object detection, semantic segmentation, and instance segmentation techniques. Our analysis includes cutting-edge AI models such as YOLOv8, DeepLabV3+, and the Real-Time Detection Transformer (RT-DETR), alongside traditional machine learning algorithms like Stochastic Gradient Descent (SGD), Decision Trees (DT), and Random Forest (RF), applied across different colour spaces. This study not only highlights the potential of AI in enhancing food sorting processes but also provides valuable insights into the selection of appropriate AI techniques for specific food industry applications.