Image Classification System Design for Yarn Inspection Process Using Teachable Machine and Raspberry Pi

This research develops a yarn classification system using machine learning with Teachable Machine and Python embedded in a Raspberry Pi-4. The system classifies four categories of yarn: Good, Striped, Moldy, and Dirty yarns. Models were created with Teachable Machine and integrated into the Raspberry Pi-4 for real-time classification. Tests showed varying inference speed and Frames per Second (FPS) for each category. The results revealed FPS and inference speed values for Good Yarn (208.05 ms, 4.46 FPS), Striped Yarn (209.73 ms, 4.42 FPS), Moldy Yarn (153.36 ms, 6.52 FPS), and Dirty Yarn (140.20 ms, 6.72 FPS). Classification accuracy was high, with Good Yarn at 97.6%, Striped Yarn at 84.8%, Moldy Yarn at 91.2%, and Dirty Yarn at 95.2%. The overall accuracy was 92%, indicating the effectiveness of the system. These findings demonstrate the system's ability for high-speed, accurate, and precise real-time yarn classification, making it feasible for the textile industry. The use of Teachable Machine and Raspberry Pi-4 offers an efficient solution to develop a machine learning-based classification system that can be applied across industries.