Unifies input and predicted meshes under a shared topological framework to enable high-fidelity 3D reconstruction with sharp features.
March 26, 2026
Original Paper
TopoMesh: High-Fidelity Mesh Autoencoding via Topological Unification
arXiv · 2603.24278
The Takeaway
Current 3D VAEs suffer from a representation mismatch that results in smoothed-out geometry and lost details. This work provides explicit mesh-level supervision, fundamentally raising the quality ceiling for 3D generative pipelines.
From the abstract
The dominant paradigm for high-fidelity 3D generation relies on a VAE-Diffusion pipeline, where the VAE's reconstruction capability sets a firm upper bound on generation quality. A fundamental challenge limiting existing VAEs is the representation mismatch between ground-truth meshes and network predictions: GT meshes have arbitrary, variable topology, while VAEs typically predict fixed-structure implicit fields (\eg, SDF on regular grids). This inherent misalignment prevents establishing explic