Learning to Rasterize Differentiably

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Chenghao, Mailee, Hamila, Montazeri, Zahra, Ritschel, Tobias
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917721805946880
author Wu, Chenghao
Mailee, Hamila
Montazeri, Zahra
Ritschel, Tobias
author_facet Wu, Chenghao
Mailee, Hamila
Montazeri, Zahra
Ritschel, Tobias
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2211_13333
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Learning to Rasterize Differentiably
Wu, Chenghao
Mailee, Hamila
Montazeri, Zahra
Ritschel, Tobias
Graphics
Computer Vision and Pattern Recognition
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.
title Learning to Rasterize Differentiably
topic Graphics
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2211.13333