One Code per Material: Neural Representation and Generation of Measured BTFs

University of Cambridge
Studio scene rendered with measured and generated BTFs

From measured to sampled BTFs in one scene, rendered with Mitsuba 3. Every textured surface is a BTF reproduced or generated by our models: the per-material networks and the compact model represent measured materials, and the flow model samples further ones; only the paintings and plants use standard textures. Close-ups: towards grazing angles, the relief of leather, fabric and carpet keeps the measured shading and shadowing between neighbouring texels, an effect a per-texel SVBRDF cannot represent.

Abstract

Measured bidirectional texture functions (BTFs) capture the full appearance of real materials—including parallax, self-shadowing and inter-reflections that analytic SVBRDFs cannot express—but each measurement is billions of values, and every new material requires a new capture. Existing neural BTF representations encode each material as large per-texel latent maps rather than as a point in a shared space from which new materials could be sampled.

We present neural representations of measured BTFs at two operating points. A per-material coordinate network outperforms the recent neural BTF of Xia et al. at matched size (+3.2 dB PSNR) and all prior baselines on every metric. A compact model represents each material by a single 256-dimensional code that conditions a shared decoder; it exceeds the PSNR (+1.2 dB) and colour accuracy of the per-texel latent maps of Rainer et al. with 256 numbers per material instead of millions, and one decoder shared by all materials.

Because each material is a single point in this code space, measured BTFs can be sampled: a class-conditional flow over the codes, trained on the 84 materials enlarged with colourways and hue rotations, generates materials that render and relight as full BTFs.

One code per material, on objects

Measured materials, each represented by a single code and decoded as a full BTF, rendered on objects with Mitsuba 3 under an outdoor environment map; the teal teapot is a hue rotation of carpet05 from our augmented dataset. Click an image for the full-resolution render.

×

Overview

Method overview

The compact model encodes each measured material and its colourways into one 256-float code with a set encoder; a decoder shared by all materials maps the code, texel position and light and view directions to RGB. A class-conditional rectified flow fitted to these codes samples new codes, which the same decoder renders as full BTFs that can be relit. Top right: measured materials reconstructed by the compact model, rendered on objects.

Reconstruction

Pick a material, a light–view pair and a method. Drag across the image to compare it with the measured BTF.

measured method
method measured
light–view pair

Nine UBO2014 materials at eight light–view pairs of the measured set; one exposure per material, the same for every method. PSNR of this image against the measured image.

Rate-distortion

Rate–distortion on the nine materials: PSNR against the number of values stored per material. The compact model is shown with its 256-float code and with the shared decoder amortised over 84 materials.

Colourways

Colourways

One measured material per row (left) and four of its colourways: its structure in the colours of other materials of its class, as flash photographs. Every value is a measured value under one 3×3 colour transform, so the colourways keep the structure, relief and shadowing of the measured BTFs.

Generation

Generated materials

Sampled materials as flash photographs: four random samples per class of each generator, without selection. The Gaussian baselines sample our code space and are decoded by our decoder; Henzler et al. and PhotoMat are retrained or fine-tuned on the same 886 training materials.

Relighting sampled materials

Every sample is a full BTF. Move the light with the slider.

light

Two random samples per class of our generator, camera fixed, lit by a moving point light.

BibTeX

@article{gokbudak2026onecode,
  title   = {One Code per Material: Neural Representation and Generation of Measured BTFs},
  author  = {Fazilet Gokbudak and Canberk Baykal and Chenliang Zhou and Rafal Mantiuk and Cengiz Oztireli},
  journal = {arXiv preprint arXiv:2611.20948},
  year    = {2026}
}