A Comparative Study on the Effect of Grid Size and Interpolation Techniques in 3D Shape Classification with CNN

This paper presents a comprehensive study on 3D shape classification utilizing Convolutional Neural Networks (CNNs). The research investigates the performance variations arising from different grid sizes and the application of interpolation techniques. A systematic analysis is conducted to determine how these factors influence the accuracy and efficiency of 3D shape classification. The experiments reveal that grid size significantly impacts the CNN's ability to accurately classify shapes with respect to computation time, with larger grid sizes enhancing classification accuracy and other performance metrics. When interpolation techniques are applied, performance metrics further improve, with the highest grid sizes achieving optimal accuracy, precision, recall, and F1-score. These findings contribute valuable insights for optimizing 3D shape classification models, demonstrating the crucial roles of both grid size and interpolation techniques in achieving high classification accuracy and efficiency in practical applications.