Learning to Rasterize Differentiably
This work addresses the need for automated softening function selection in differentiable rendering, which is incremental as it builds on prior analyses of softening combinations.
The paper tackles the problem of selecting optimal softening functions for differentiable rasterization, which is crucial for gradient flow in rendering, by parameterizing the continuous space of common softening operations and meta-learning tunable softness functions. The result is a method that generalizes to new and unseen differentiable rendering tasks, such as 2D and 3D shape, pose, and occlusion, with optimal softness, though no concrete numbers are provided.
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.