We propose a novel reflectance decomposition network that can estimate shape, BRDF, and per-image illumination given a set of object images captured under ...

Our neural-PIL takes as input a latent vector representation for the environment map, the surface roughness, and an incident ray direction, and from them ...

A novel method which decomposes multiple images into shape, BRDF and illumination with a split-sum preintegrated illumination network.

We propose a novel reflectance decomposition network that can estimate shape, BRDF, and per-image illumination given a set of object images captured under ...

Nov 9, 2021 · A method that decomposes multiple images of a scene into shape, reflectance and illumination using a neural illumination model.

We train the neural-PIL using the same environment maps dataset as used for training the Light-SMAE. Here, the encoder of the Light-SMAE is used for defining ...

Oct 27, 2021 · Our key technique is a novel illumination integration network called Neural-PIL that replaces a costly illumination integral operation in the ...

We propose a novel reflectance decomposition network that can estimate shape, BRDF, and per-image illumination given a set of object images captured under ...

Neural-PIL [21] explored to approximate pre-integration light integration process by a neural network [65] to achieve fast calculation of outgoing radiance.

Decomposing a scene into its shape, reflectance and illumination is afundamental problem in computer vision and graphics. Neural approaches such asNeRF have ...