Neural Differential Radiance Field: Learning the Differential Space Using a Neural Network

We introduce an adjoint-based inverse rendering method using a Neural Differential Radiance Field, i.e. a neural network representation of the solution of the differential rendering equation. Inspired by neural radiosity techniques, we minimize the norm of the residual of the differential rendering equation to directly optimize our network. The network is capable of outputting continuous, view-independent gradients of the radiance field w.r.t scene parameters, taking into account differential global illumination effects while keeping memory and time complexity constant in path length. To solve inverse rendering problems, we simultaneously train networks to represent radiance and differential radiance, and optimize the unknown scene parameters. Our method is not scalable to millions of scene parameters, but we propose future work directions that could make that happen in the future.