SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation

2026-10-01

AI Summary

This study addresses the high computational overhead of voxel tokens and weak topological consistency in high-resolution 3D generation by proposing the SILSA framework. The method leverages sliding-window slice latents to enable single-stage rectified flow generation, maintaining surface continuity through a compact representation. It further introduces a novel slice-level topological supervision mechanism that ensures structural correctness by matching persistence diagrams and aligning Betti number transitions. By integrating techniques such as a slice VAE, a sparse volumetric decoder, and a volumetric anchor lattice, the proposed approach achieves substantial improvements. Experimental results demonstrate an 8.7% increase in PSNR, a 5.96-point gain in coverage, a 9.2% reduction in Betti error, a 40.4% decrease in memory consumption, and a 58.5% acceleration in inference speed.

Abstract

High-resolution 3D generation increasingly relies on voxel latents and multi-stage pipelines that first predict active structure and then synthesize local geometry. While effective, this design fragments continuous surfaces into many local tokens, inflates generation cost, and often weakens topological consistency for thin or highly connected shapes. We introduce SILSA, a topology-aware 3D generation framework that represents shapes with compact sliding-window slice latents. Instead of generating expensive voxel tokens, SILSA uses a fixed set of overlapping slices along the three canonical axes, where each token summarizes a local depth window to preserve cross-sectional continuity and support single-stage rectified-flow generation. A Slice VAE encodes oriented surface samples into multi-axis slice latents and reconstructs them with a sparse volumetric decoder, while a Volumetric Anchor Lattice coordinates directional slice streams through a shared 3D workspace. To preserve structural correctness, we introduce slice-level topology supervision that matches persistence diagrams and aligns Betti transitions across neighboring slices. Experiments show that SILSA improves structural fidelity while substantially reducing generation cost. SILSA improves PSNR by $8.7\%$, coverage by $5.96$ absolute points, and Betti error by $9.2\%$ over the strongest baseline, while using $70.0\%$ fewer tokens than the next-most compact baseline and over $98\%$ fewer tokens than sparse or hierarchical tokenizers, effectively reducing training memory by $40.4\%$ and inference time by $58.5\%$. Qualitative results further show improved preservation of thin structures, repeated components, and long-range connectivity.

Problem

• 3D generation
• topology preservation
• voxel latents
• generation cost
• structural fidelity

Innovation

• sliding-window slice latents
• topology-preserving generation
• Slice VAE
• volumetric anchor lattice
• Betti transitions supervision