AI Paper Atlas

Research map / Image generation / GANs

SurfaceNet: Adversarial SVBRDF Estimation From a Single Image

Giuseppe Vecchio, Simone Palazzo, Concetto Spampinato

ICCV 2021 · Image generation › GANs · approach: GAN-based image manipulation · 47 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add learnable upsampling decoder to the base generator (ablation)SurfaceNet generator · Synthetic SVBRDF (flash)rendering RMSE: 0.185 -> 0.087Improved
add skip connections with downsampled input after decoder (ablation)SurfaceNet generator · Synthetic SVBRDF (flash)rendering RMSE: 0.087 -> 0.073Improved
add skip connections (ablation)SurfaceNet generator · Synthetic SVBRDF (flash)normal RMSE: 0.093 -> 0.122Hurt
add image-level discriminator with adversarial loss (ablation)SurfaceNet generator · Synthetic SVBRDF (flash)rendering RMSE: 0.073 -> 0.067Improved
use patch-based instead of image-based discriminator (ablation)SurfaceNet · Synthetic SVBRDF (flash)rendering RMSE: 0.067 -> 0.058Improved
use patch discriminator instead of image discriminator (ablation)SurfaceNet · Synthetic SVBRDF (flash)roughness RMSE: 0.065 -> 0.029Improved
add MS-SSIM to L1 supervised loss (ablation)SurfaceNet · Synthetic SVBRDF (flash)rendering RMSE: 0.134 -> 0.088Improved
add adversarial loss to supervised loss (ablation)SurfaceNet · Synthetic SVBRDF (flash)rendering RMSE: 0.088 -> 0.058Improved
use adversarial loss alone instead of full loss (ablation)SurfaceNet · Synthetic SVBRDF (flash)diffuse RMSE: 0.017 -> 0.133Hurt
include unlabeled real images in training (ablation)SurfaceNet · Synthetic SVBRDF (flash)rendering RMSE: 0.058 -> 0.064Hurt
include unlabeled real images in training (ablation)SurfaceNet · Synthetic SVBRDF (flash)roughness RMSE: 0.029 -> 0.043Hurt
include unlabeled real images in training (ablation)SurfaceNet · Real photographsqualitative SVBRDF qualityImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy