Wong, Chee-Kien GabriyelWang, JianliangDieter Fellner and Charles Hansen2015-07-192015-07-192006https://doi.org/10.2312/egs.20061035The real-time rendering process is well known to be extremely dynamic and complex. This paper presents a novel approach to modeling this process via the system identification methodology. Given the process s dynamic nature arising from the possible myriad variations of render states, polygon streams and the non-linearities involved, we describe a modeling approach using neural networks with supervised training from application-generated data. By comparing the outputs of the neural network model s representation of the rendering process with actual empirical data, we discuss the accuracy of our approach in relation to the practical issues of integrating this study to real-world applications.10.2312/egs.2006103589-93