Distributions for Compositionally Differentiating Parametric Discontinuities

Jesse Michel, Kevin Mu, Xuanda Yang, Sai Praveen Bangaru, Elias Rojas Collins, Gilbert Bernstein, Jonathan Ragan-Kelley, Michael Carbin, Tzu-Mao Li

Introduces a new language (Potto) that uses distribution theory to resolve some of the shortcomings of the Teg language, and provides a fully modular approach to differentiating & composing applications with discontinuities.

Importance Sampling BRDF Derivatives

Yash Belhe, Bing Xu, Sai Praveen Bangaru, Ravi Ramamoorthi, Tzu-Mao Li

Exploits common structure across a variety of BRDF models to derive efficient importance sampling methods for their derivatives, through a mix of positivization and product/mixture decomposition

SLANG.D: Fast, Modular, and Differentiable Shader Programming

Sai Praveen Bangaru, Lifan Wu, Tzu-Mao Li, Jacob Munkberg, Gilbert Bernstein, Jonathan Ragan-Kelley, Fredo Durand, Aaron Lefohn, Yong He

HLSL-like shading language that supports first-class automatic differentiation and object-oriented programming for large-scale differentiable rendering frameworks and inter-operation with tensor ML frameworks

Warped-Area Reparameterization of Differential Path Integrals

Peiyu Xu, Sai Praveen Bangaru, Tzu-Mao Li, Shuang Zhao

Received a Best Paper Award.   

This paper shows how to combine the warped-area reparameterization with the material-form parameterization to present an efficient way to estimate derivatives w.r.t geometry in the presence of complex multi-bounce illumination.

Differentiable Rendering of Neural SDFs through Reparameterization

Sai Praveen Bangaru, Michael Gharbi, Tzu-Mao Li, Fujun Luan, Kalyan Sunkavalli, Milos Hasan, Sai Bi, Zexiang Xu, Gilbert Bernstein, Fredo Durand

This work shows how to differentiate a neural SDF renderer without approximations using reparameterization, with applications to 3D reconstruction on real and synthetic multi-view datasets.

Systematically Differentiating Parametric Discontinuities

Sai Praveen Bangaru*, Jesse Michel*, Kevin Mu, Gilbert Bernstein, Tzu-Mao Li, Jonathan Ragan-Kelley

This work systematically address the challenges of differentiating parametric discontinuities under integration, by developing a programming language, proving correctness of its semantics, and demonstrating applications including shader design and contact simulation.

Unbiased Warped-Area Sampling for Differentiable Rendering

Sai Praveen Bangaru, Tzu-Mao Li, Fredo Durand

First paper to derive and implement an unbiased estimator in the area-measure for the differentiable rendering equation.

Towards Reflectometry from Inter-reflections

Kfir Shem-Tov*, Sai Praveen Bangaru*, Anat Levin, Ioannis Gkioulekas

Applies a path-space differentiable rendering algorithm to improve photometric BSDF recovery using concave objects.

Towards Shape Reconstruction through Differentiable Rendering

Sai Praveen Bangaru

A solution to the problems with shape-recovery methods. Uses a mitsuba shape-differentiable path tracer and a TensorFlow optimizer to obtain accurate reconstructions in the presence of interreflections and non-lambertian surfaces.

Action Conditional Projection Neural Networks

Sai Praveen Bangaru

Similar to Action-Conditional Video Prediction, which predicts the next frame of an ATARI game given the user input, ACPNNs are designed to predict 2D and 3D environments where the output is based on perspective vision.

Multi-task Reinforcement Learning

Sai Praveen Bangaru, Suhas Jayaram, Balaraman Ravindran

Exploration in multi-task reinforcement learning is critical in training agents to deduce the underlying MDP. We present a novel method to facilitate exploration in multi-task reinforcement learning using deep generative models.