Learning to Rasterize Differentiably

TL;DR

This paper introduces a learnable framework for differentiable rasterization that optimizes softness functions via meta-learning, improving generalization across various inverse rendering tasks.

Contribution

It proposes parameterizing the space of softening functions and meta-learning their optimal parameters for better performance and generalization in differentiable rendering.

Findings

Meta-learned softness functions outperform fixed choices.

Improved convergence and accuracy in inverse rendering tasks.

Generalizes well to unseen rendering scenarios.

Abstract

Differentiable rasterization changes the standard formulation of primitive rasterization -- by enabling gradient flow from a pixel to its underlying triangles -- using distribution functions in different stages of rendering, creating a "soft" version of the original rasterizer. However, choosing the optimal softening function that ensures the best performance and convergence to a desired goal requires trial and error. Previous work has analyzed and compared several combinations of softening. In this work, we take it a step further and, instead of making a combinatorial choice of softening operations, parameterize the continuous space of common softening operations. We study meta-learning tunable softness functions over a set of inverse rendering tasks (2D and 3D shape, pose and occlusion) so it generalizes to new and unseen differentiable rendering tasks with optimal softness.