Differender

Taichi-based Differentiable SciVis Renderer for PyTorch

Features & Status

• Differentiable Rendering
  • Differentiating on transfer functions (TFs)
  • Backpropagation (BP) on transfer functions with a momentum optimizer
  • Differentiating and BP on volume
  • Advance loss for compensating high frequency signals
• Direct Volume Rendering
  • Ray jittering
• Engineering
  • Advance data layout for speedup
  • Logging on TFs
  • PyTorch integration