A special issue tracking how neural networks are rewriting the pipeline for meshes — from implicit-to-explicit reconstruction and generative synthesis, to topology-aware learning and differentiable rendering.
Three-dimensional meshes remain the working representation across scientific visualization, graphics, VR/AR, and digital-twin pipelines — but deep learning has changed almost everything about how they're produced and consumed. Neural implicit fields (SDFs, occupancy networks, NeRF-style representations) now sit upstream of explicit geometry; diffusion and transformer-based generators synthesize shapes directly from text or images; and mesh-native architectures increasingly operate on vertices, edges, and faces without abandoning connectivity for point clouds or voxels.
This special issue is dedicated specifically to meshes as a representation — the irregular connectivity, topology preservation, and geometric consistency problems that point-cloud and implicit-focused venues tend to skip past. It welcomes original theory, new methods, and applied work that pushes mesh understanding forward.
Download the PDF version of this call for papers ↓
Consult the call for papers on the journal website here.
We solicit original contributions addressing theoretical and practical issues for 3D meshes, including — but not limited to.
Signed distance fields, occupancy, and implicit functions for 3D geometry.
Generation, completion, and reconstruction of mesh geometry.
Learning-based approaches to lighter, cleaner, better-conditioned meshes.
Mesh segmentation, classification, and correspondence learning.
Neural approaches to reshaping, rigging, and animating mesh geometry.
Large-scale datasets and benchmarks for mesh-focused deep learning.
GANs, diffusion, and auto-regressive models for 3D mesh synthesis.
Differentiable rendering and inverse graphics for mesh supervision.
Network designs that respect and exploit mesh connectivity.
Mesh learning driven by text, images, or video.
Mesh quality metrics, and fairness/bias in 3D data generation.
Visualization, simulation, AR/VR, and industrial design applications.
Submitted papers are screened by all guest editors for fit and originality, then distributed by topic and expertise. Only one round of major/minor revision is allowed to keep the schedule on track.
Manuscripts should follow the standard author guidelines of Image and Vision Computing and be flagged for this special issue at submission.
Details on the journal can be found here.
Guidelines for authors can be found here.
Select the special issue article type for “3D Mesh Processing and Generation in the Age of Deep Learning” when submitting, and mention the special issue in your cover letter.