|
Enhancing object classification accuracy by incorporating tangent data in the model architecture Cristina Popirlan, Irina-Valentina Tudor, Cristina-Mihaela Tudorache, Claudiu-Ionut Popirlan, Constantin-Cristian Dinu, Daniela Danciulescu, Mihaela Colhon, Gabriel Stoian |
1-24 |
|
Enhancing object classification accuracy by incorporating tangent data in the model architecture Dragos-Bogdan Tudor, Elena Pelican |
25-36 |
1 University of Craiova, Department of Computers and Information Technology
Blvd. Decebal nr. 107, RO-200440, Craiova, Romania
2 University of Craiova
Craiova, Romania
Abstract. As the integration of gamification elements in educational apps continues to gain prominence, there is a growing need for effective evaluation methods to assess their impact on learning outcomes. This article presents a questionnaire-based approach for evaluating students' depth of knowledge and perception about the existence of gamification elements in apps, as well as extracting the most appropriate elements based on the games types the respondents play in their everyday lives. The questionnaire, which was tested for consistency and reliability, includes a comprehensive set of questions that capture various dimensions of gamification, including game mechanics, feedback systems, progress tracking, rewards, and social interaction. We performed several analyses on the collected answers and the obtained results led us to the conclusion that gamification elements are worth including in the design of future educational applications. The resulting improvements we suggest can enable the creation of a more effective and appealing learning experience by enhancing student engagement and motivation.
Keywords: gamification, education, gamified apps, questionnaire study, smart learning, active learning methods, learning outcomes, AI
Cite this paper as:
Popirlan, C., Tudor, I.-V., Tudorache, C.-M., Popirlan, C.-I., Dinu, C.-C., Danciulescu, D., Colhon, M., Stoian, G. Enhancing object classification accuracy by incorporating tangent data in the model architecture.
International Journal of User-System Interaction 17(1),
1-24, 2024.
1 Ovidius University of Constanţa
124 Mamaia Blvd., 900527, Constanta, Romania
2 Ovidius University Constanta
124 Mamaia Blvd., Constanta, Romania
Abstract. This paper presents a method that aims to improve the recognition rate of PointMLP (Xu et al., 2022) for the classification task with minimal impact on training and inference speed by extracting and incorporating more information from the 3D objects used during the training process. Firstly, when extracting the 3D point cloud from the 3D mesh, besides sampling 3D points across the surface of the 3D objects and applying other common preprocessings, our proposed algorithm also saves the normal of the triangle out of which the 3D point was sampled and calculates 2 orthogonal vectors that are tangent to the surface of the 3D model in that point. Afterwards, the resulting vectors are incorporated into the architecture of PointMLP through the addition of new layers to the calculation of the embedding vector, so that the information about the tangents and normals is incorporated into the model architecture. This effectively provides the model with more context about the surface of the original 3D object, slightly improving the accuracy of the model with close to 0 impact on training/inference speed.
Keywords: 3D objects classification, PointMLP framework, tangents/bi-tangents/normals
Cite this paper as:
Tudor, D.-B., Pelican, E. Enhancing object classification accuracy by incorporating tangent data in the model architecture.
International Journal of User-System Interaction 17(1),
25-36, 2024.