AI Paper Atlas

Research map / 3D vision / Face and shape

Towards High Fidelity Monocular Face Reconstruction With Rich Reflectance Using Self-Supervised Learning and Ray Tracing

Abdallah Dib, Cédric Thébault, Junghyun Ahn, Philippe-Henri Gosselin, Christian Theobalt, Louis Chevallier

ICCV 2021 · 3D vision › Face and shape · approach: Face reflectance and relightable rendering · 58 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
use 9-band spherical harmonics instead of 3-band (ablation)ray-traced face reconstruction · Face reconstruction (qualitative)shadow estimation qualityImproved
use 9-band spherical harmonics instead of 7-band (ablation)ray-traced face reconstruction · Face reconstruction (qualitative)shadow estimation quality: slightly betterImproved
use two separate networks for diffuse and specular increments instead of one (ablation)albedo increment estimation · Face reconstruction (qualitative)albedo separationImproved
add symmetry and consistency regularizers (ablation)albedo personalization · Face reconstruction (qualitative)residual shadows in albedoImproved
train encoder jointly with the two albedo networks (ablation)two-stage training · Face reconstruction (qualitative)geometry fitImproved
use differentiable ray tracing instead of a vertex-based renderer (ablation)face reconstruction · Face reconstruction (qualitative)shadow qualityImproved
increase ray tracing samples per pixel from 8 to 16 or 32 (ablation)ray-traced training · trainingreconstruction quality: 8, 16 and 32 sppNo clear effect
model self-geometry ray bounces (ablation)ray-traced training · trainingaccuracyNo clear effect
use predicted attributes as initialization for optimization-based methodself-supervised network · Face reconstruction (qualitative)reconstruction qualityImproved
ours versus best prior method on vertex position errorOurs · Geometry error on 23 GT-mesh imagesvertex position error (cm): 0.174 vs 0.181Mixed
ours versus a noisy-mesh prior method on normal errorOurs · Geometry error on 23 GT-mesh imagesnormal error (rad): 0.159 -> 0.148Improved

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