This is a fork of tiny-cuda-nn. See the original repository for details.

Overview

This repository implements a neural model to encode material properties on the surfaces of 3D objects as described in the thesis Neural Compression of Material Properties using a Geometry-Associated Feature Hierarchy. It builds on tiny-cuda-nn, a fast framework to train and evaluate small MLPs. We contribute a novel input encoding as described in the thesis as well as an example.

Usage

neural_surface <path_to_object_basedir> <object_filename> <sample_path>

• path_to_object_basedir is the filepath to the directory containing the .obj, .mtl and texture files
• object_filename is the filename of the .obj file in the base directory
• sample_path is the path to a .csv file containing a list of surface positions to validate the model

The validation file is structured as follows:

image_width, image_height 		// Width and height of the validation image
triangle_id, t_0, t_1			// Each row represents a pixel in the image and contains the surface position
triangle_id, t_0, t_1			// 		 that the pixel lands on (t_2 is implicit)
...