MRPNN - A neural volumetric renderer

About

MRPNN is a neural volumetric renderer that can render high-order volumetric scattering in real time. See our paper for more details. SIG23: Deep Real-time Volumetric Rendering Using Multi-feature Fusion

Project page | Supplementary doc | Dataset for demo

Install

Requirements

• CUDA 11.x

  Version	Compatible	Info
  11.0	Yes*	CUDA11.0 Only support sm_75
  11.1	Yes	Recommended
  11.2	Yes	Recommended
  11.3	Yes*	11.3 update 1 fix a compile error
  11.4~11.7	Not Sure
  11.8	Yes*	After commit 324eae2241bfb51cae606f4e626d48ce1e0afdca

• Glfw3 and GLEW if you want to compile with GUI
• x64 Ubuntu or Windows
• Nvidia GPU that supports sm_75 or sm_86

Test with Geforce RTX2000 series, RTX3000 series, RTX4000 series, and Quadro A5000 GPU.

To build on Linux:

• Install CUDA 11.x.
• Unzip 'tools/curand include.rar' into your cuda include directory.
• Run CMake to generate the project, and use -T cuda=<PATH/TO/CUDA/toolkit> to select a cuda version.
• If you are using the RTX3000 series (or another sm_86 card), enable the RTX30XX option.
• Enable the RPNN option if you want to compare with it.
• Compile your project, it may stuck for a while.
• Linux currently doesn't support build with GUI.

To build on Windows:

• Install CUDA 11.x.
• Install glew and glfw3 if you need GUI support. (Remember to install the x64 version)
• Run CMake to generate the project, and use -T cuda=<PATH/TO/CUDA/toolkit> to select a cuda version.
• If you are using CUDA 11.0~11.3, you may also need to have MSVC v14.25 and using -T version=14.25.
• If you are using the RTX3000 series, enable the RTX30XX option.
• Enable the RPNN option if you want to compare with it.
• Check the GUI option if needed.
• Compile your project, it may stuck for a while.

How to use

Use test case

• Download test data from OneDrive.
• Unzip them to ./TestCase/.
• Run the Test.

The bias comparison is only supported on Windows!

Use GUI

• Download Cloud0.rar.
• Unzip it to ./TestCase/.
• Build project with GUI option on.
• Run the VolumeRender.

Use custom data

VolumeRender v(/*volumeric data resolution*/512);

// filling volume with a sphere centered at (0.5,0.5,0.5)
v.SetDatas([](int x, int y, int z, float u, float v, float w) {
    float dis = distance(make_float3(0.5f, 0.5f, 0.5f), make_float3(u,v,w));
    return dis < 0.25 ? 1.0f : 0;
    });

// Call Update after changing volumetric data.
v.Update();

// Call render to generate result in memory (array of float3 colors)
v.Render(...)

// or create a camera of this volume
Camera cam(v);

// and use camera to render to file
cam.RenderToFile(...)

BibTex

@inproceedings{ MRPNN:SIG:2023,
	author       = {Jinkai Hu and Chengzhong Yu and Hongli Liu and Ling-qi Yan and Yiqian Wu and Xiaogang Jin},
	title        = {Deep Real-time Volumetric Rendering Using Multi-feature Fusion},
	booktitle    = {{SIGGRAPH} '23: Special Interest Group on Computer Graphics and Interactive Techniques Conference, Los Angeles CA, United States, August 6 - 10, 2023},
	publisher    = {{ACM}},
	year         = {2023}
}