Deep Learning Classification in web3D model geometries

What is it about?

Web3D appears to be very important for the future, especially after the positive prospects and expectations of Metaverse. However, Web3D is a special case of 3D models designed for web applications, thus should be small, lightweight with relatively low level of detail and consequently more difficult to classify. In this work we use Deep Learning, a new generation of multilayer Neural Network algorithms able to classify objects by processing at their geomerty, for real time predictions in Web3D meshes on a webpage. Also we provide methods for Wed3D data preprocessing, to achieve more accurate prediction.

Featured Image

Photo by h heyerlein on Unsplash

Photo by h heyerlein on Unsplash

Why is it important?

In our experiments we prove that web3D models with some real time geometrical enhancement may be processed with Deep Learning algorithms with a high rate of positive prediction results. Moreover, we demonstrate that a certain family of Deep Learning algorithms may efficiently classify 3D models created by scanner, camera or CAD at the same time with the same training.