AI Paper Atlas

Research map / 3D vision / Face and shape

Parallelised Differentiable Straightest Geodesics for 3D Meshes

Hippolyte Verninas, Caner Korkmaz, Stefanos Zafeiriou, Tolga Birdal, Simone Foti

CVPR 2026 · 3D vision › Face and shape · approach: Learned mesh and geometric representations · 0 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 GPU-parallel straightest geodesics instead of CPU implementationsExponential map · large batch sizesruntime: 3 orders of magnitude fasterImproved
use GFD instead of EP differentiation (ablation)Exp backward pass · sphere gradient checkgradient accuracy: backward about 3.85 times more expensive than EPMixed
keep gradients w.r.t. start point p in EP (ablation)Extrinsic proxy · sphere gradient checkgradient accuracyHurt
use our Exp instead of projection integration (PI)Exponential map · high face count meshesruntime: 5 to 6 orders of magnitude fasterImproved
use learned patch size via differentiable Exp (AGC) instead of fixed-patch GCNNGCNN · body part segmentationaccuracy: 86.4% -> 91.7% (XYZ), 92.3% (HKS)Improved
use HKS input instead of XYZ coordinates (ablation)AGC · body part segmentationaccuracy: 91.7% -> 92.3%Improved
use GFD instead of EP for AGC training (ablation)AGC · body part segmentationaccuracy and training time: 1 min 21 s -> 2 min 8 s per epochNo clear effect
use MeshFlow with exponential map instead of RFMRiemannian flow matching on meshes · Bunny/Spot eigenfunctionsinference speed and memory: 1.6e4 times faster, 97% less GPU memoryImproved
use MeshFlow instead of RFMRiemannian flow matching on meshes · Bunny/Spot eigenfunctionsNLL: 1.55 -> 1.71 (Bunny k=100)Hurt
use more samples (1024 vs 256) for OT coupling (ablation)MeshFlow · Bunny k=100KLD: 0.13 -> 0.058Improved
use Mesh-LBFGS instead of Lloyd's algorithmGeodesic centroidal Voronoi tessellation · Spot the Cow, 50 seedsconvergence and final energyImproved
start seeds clustered rather than uniform (ablation)Mesh-LBFGS · Spot the Cow, 50 seedsadvantage over LloydImproved

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