Learning adaptive sampling and reconstruction for Volume Visualization

Currently supported:

• Iso-surface rendering with screen-space shading
• Direct Volume Rendering
• Temporal consistency and reprojection
• adaptive sampling with an importance map in screen space
• adaptive sampling in object space by changing the step size

Demonstration: https://sites.google.com/view/sebastian-weiss/research/adaptive-sampling

Project structure:

• renderer: a shared library exposing PyTorch operation that contains the rendering core (C++, CUDA)
• network: super-resolution network training and testing code (Python, PyTorch)
• inference-gui: interactive gui combining the renderer and networks, allows to test all available options (C++, OpenGL)

See the release page for binaries, datasets and pretrained networks

Requirements

• CUDA >= 1.1
• Python >= 3.6
• PyTorch >= 1.5
• OpenGL

Tested with CUDA 10.1, Python 3.6, PyTorch 1.5, Windows 10 and Ubuntu 18