Real-time photo-realistic clouds path tracer with Lorenz-Mie scattering.

This is an enchancement on paper by Disney Research Deep Scattering.

Published and presented at SIGGRAPG Asia

Preparation

• Generate a sample cloud in Houdini
• Exporter of volumetric data from Houdini
• Importer of volumetric data for python and c++

Ground Truth

• Naive implementation of path tracer that only takes into account Beer-Lambert law and direct scattering
• Interactive visualisation of results
• Naive implementation that also takes into account Lorenz-Mie phase function.
• Implement multiple scattering using Monte-Carlo method
• Implement progressive path tracing
• Implement custom progressive path
• Implement Reinhard tone mapping
• Implement next event estimation
  

Dataset

• Generate more clouds
• Gather dataset of (cloud_id, point_in_cloud, sun_direction, view_direction, radiance)
• Gather dataset of samples around the point according to the original paper
• Separate the dataset for training, test, validation with total size of 15`000`000; leave a few clouds out of those sets for visual evaluation

Training

• Build a model in pytorch as described in the original paper
• Train the model
• Implement an efficient neural-based renderer in c++
• Compare to the original results

  

Research Experiments with different architectures.

• Shadow maps
• Latent space light probes
• Light probe rotation
• Light probe interpolation (trilinear, barycentric)
• Evaluate radiance only for the outer layers of the cloud