MINSU: Precision Quantity Counter with DNN-based Volume Estimation

This paper introduces MINSU, the first DNN (Deep Neural Network) based quantity counter specifically designed for smart factories. The innovative DNN design of MINSU positions it as a suitable solution for the emerging trend of fully automated, man-less smart factories. By periodically updating inventory status using live images, MINSU eliminates the need for manual, repetitive inventory-checking tasks. This system is easily adaptable to existing industrial environments, as it does not require the installation of new racks or modifications to the current factory floor plan. MINSU is designed to seamlessly navigate within the existing factory setup, achieving a level of performance that surpasses manual inventory counting methods. We have tested MINSU in a lab environment replicating a small-sized factory, using actual bolts as the target inventory. Our experiments demonstrated an accuracy of 84.33%, showcasing the potential for real-world applications in modern smart factories.