A novel deep ordinal classification approach for aesthetic quality control classification

Abstract Nowadays, decision support systems (DSSs) are widely used in several application domains, from industrial to healthcare and medicine fields. Concerning the scenario, we propose a DSS oriented aesthetic quality control (AQC) task, which has quickly established itself as one of most crucial challenges Industry 4.0. Taking into account increasing amount data this domain, machine learning (ML) deep (DL) techniques offers great opportunities automatize overall AQC process. State-of-the-art is mainly approach problem with nominal DL classification method does not exploit ordinal structure thus penalizing error among distant classes (which relevant aspect for real use case). The paper introduces methodology classification. Differently other methods, combined standard categorical cross-entropy cumulative link model imposed constraint via thresholds slope parameters. Experimental results were performed solving an task on novel image dataset originated specific company’s demand (i.e., assessment wooden stocks). We demonstrated how proposed able reduce misclassification errors (up 0.937 quadratic weight kappa loss) while overcoming state-of-the-art models reducing bias factor related item geometry. was integrated main core supported by Internet Things (IoT) architecture that can human operator up 90% time needed qualitative analysis carried out manually domain.