Plateau-reduced Differentiable Path Tracing
This enables optimization of intricate light transport problems like caustics or global illumination that existing differentiable renderers fail on, benefiting computer graphics and vision researchers.
The paper tackles the problem of inverse rendering not converging due to plateaus in the objective function by convolving the rendering function with a kernel to blur the parameter space, resulting in net-gains in optimization error and runtime performance.
Current differentiable renderers provide light transport gradients with respect to arbitrary scene parameters. However, the mere existence of these gradients does not guarantee useful update steps in an optimization. Instead, inverse rendering might not converge due to inherent plateaus, i.e., regions of zero gradient, in the objective function. We propose to alleviate this by convolving the high-dimensional rendering function that maps scene parameters to images with an additional kernel that blurs the parameter space. We describe two Monte Carlo estimators to compute plateau-free gradients efficiently, i.e., with low variance, and show that these translate into net-gains in optimization error and runtime performance. Our approach is a straightforward extension to both black-box and differentiable renderers and enables optimization of problems with intricate light transport, such as caustics or global illumination, that existing differentiable renderers do not converge on.