Automated quality control by AI-driven visual inspection techniques in the food processing industry

AI-driven visual inspection techniques in industrial food processing are critical for quality assurance and safety; therefore, the selection of the most suitable algorithm should be based on both operational and performance criteria. This study tests the convolutional neural network YOLOv8 on an industrial bakery dataset under typical shop-floor stress conditions. The algorithm performance is evaluated across high-volume and resource-constrained scenarios, by means of decision-oriented performance metrics: Risk-Weighted Total Cost of Quality, Throughput Efficiency, and Waste Reduction. Results show the robustness of YOLOv8 in detecting relevant defects, ease of use and economic sustainability, thus enabling its implementation and operational effort.