Suolin Zhang

Training Neural Materials For Real-Time Rendering A Replication with INT8 Quantisation and ACF-Guided Synthesis

Summary

This project is a replication for SIGGRAPH Conference Papers ‘25 * Towards Comprehensive Neural Materials: Dynamic Structure-Preserving Synthesis with Accurate Silhouette at Instant Inference Speed*. The Github repo includes a neural network which you can use to train your own btf material and export the training features. After training, Falcor(Nvdia’s real-time redering framwork) can be used to do a real-time rendering for the exported feature bundle. Details about how to train and render are in demo. This project represents a detailed exploration and attempt into the latest rendering technology.

Thesis Training Neural Materials For Real-Time Rendering: A Replication with INT8 Quantisation and ACF-Guided Synthesis

Contact

LinkedIn: www.linkedin.com/in/suolin-zhang-391938207 email: zsl980614@gmail.com