AI Paper Atlas

Research map / 3D vision / Face and shape

Geometry Distributions

Biao Zhang, Jing Ren, Peter Wonka

ICCV 2025 · 3D vision › Face and shape · approach: Learned mesh and geometric representations · 8 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 proposed network instead of hashing grid (ablation)Geometry Distributions · Loong shapeChamfer x10^3: 4.133 -> 2.140Improved
use proposed network instead of DeepSDF-style MLP (ablation)Geometry Distributions · Loong shapeChamfer x10^3: 2.647 -> 2.140Improved
increase training set size from 10^5 to approx. infinity (resampling) (ablation)Geometry Distributions · jellyfish shapeChamfer x10^3: 4.696 -> 2.916Improved
increase training set from 10^7 to 10^8 samples (ablation)Geometry Distributions · jellyfish shapeChamfer x10^3: 2.936 -> 2.917No clear effect
increase sampling steps from 8 to 64 (ablation)Geometry Distributions · Archimedes shapeChamfer x10^3: 9.803 -> 2.780Improved
increase sampling steps from 16 to 64 (ablation)Geometry Distributions · Archimedes shapeChamfer x10^3: 2.814 -> 2.780No clear effect
increase network blocks from 2 to 10 (ablation)Geometry Distributions · lamp shapeChamfer x10^3: 3.784 -> 3.544Improved
increase network blocks from 4 to 10 (ablation)Geometry Distributions · lamp shapeChamfer x10^3: 3.565 -> 3.544No clear effect
use Gaussian instead of uniform initial noise (ablation)Geometry Distributions · lion shapeChamfer x10^3: 3.734 -> 3.246Improved
use Gaussian instead of uniform initial noise (ablation)Geometry Distributions · lamp shapeChamfer x10^3: 3.574 -> 3.538Improved
increase inversion steps from 4 to 64 (ablation)Geometry Distributions · mouse shapeinversion MSE: 6.88e-1 -> 1.76e-6Improved
model surface as distribution instead of vector-field based methodvector field methods · uniform surface samplingChamfer x10^3Improved

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