lod generation

This work addresses the challenge that modern 3D reconstruction and generation methods often yield dense, noisy, and non-manifold meshes, which are ill-suited for applications such as simulation and AR/VR that demand efficient and reliable geometry. To tackle this, the authors propose a feature-aware quadric error metric (FA-QEM) simplification framework that integrates geometric deviation, boundary curvature, and normal consistency into a multi-objective quadratic error function. Coupled with optimal vertex placement and a continuous texture mapping transfer strategy, the method effectively preserves sharp geometric features while achieving substantial mesh simplification. Experiments demonstrate that FA-QEM significantly reduces geometric error, enhances visual fidelity, and improves computational efficiency on both AI-generated and real-world datasets, exhibiting strong robustness and practical utility.

Real-time simplification of large-scale polygonal meshes faces an inherent trade-off between computational speed and geometric fidelity. Method: We propose a curvature-guided simplification algorithm based on reverse refinement, departing from conventional top-down simplification paradigms. Inspired by splitting strategies in vector quantization, our approach begins with a coarse initial approximation and progressively refines it via curvature-driven hierarchical subdivision coupled with rigorous error control. This enables guaranteed output under time constraints and high-fidelity rendering at interactive frame rates. Contribution/Results: Experimental evaluation on ultra-large-scale models demonstrates significant improvements over state-of-the-art methods: our algorithm generates shape-preserving, low-distortion approximations within single-frame milliseconds—achieving unprecedented balance between efficiency and geometric accuracy.

Addressing the challenge of simplifying non-manifold, multiply-connected, and textured triangle meshes, this paper formulates mesh simplification as a 2-dimensional simplicial complex reduction problem—the first such formulation—and introduces a topology-robust edge-collapse framework. Key contributions include: (1) an enhanced quadric error metric adapted to topological changes, ensuring geometric fidelity; (2) a novel texture simplification paradigm that retains only texture colors while decoupling UV layout optimization, thereby eliminating bleeding artifacts entirely; and (3) a color-space-driven texture remapping strategy. The method supports arbitrary topology and level-of-detail (LOD) generation. Comprehensive qualitative and quantitative evaluations, together with user studies, demonstrate consistent superiority over state-of-the-art approaches—significantly improving simplification quality and visual consistency for non-manifold meshes.

Existing generative 3D models—particularly single-view or sparse multi-view approaches—are computationally expensive and often produce coarse shapes with insufficient geometric detail. Method: We propose a lightweight, data-driven latent-space streaming enhancement framework that establishes the first “coarse-to-fine” geometric refinement paradigm. Our method jointly optimizes local detail synthesis and global structural consistency via data-dependent latent-space flow modeling, and introduces a learnable token-matching mechanism to explicitly enforce spatial correspondence. Crucially, it requires no fine-tuning of the upstream generator and achieves broad compatibility across diverse 3D generation backbones with minimal training overhead. Contribution/Results: Experiments demonstrate substantial improvements in surface fidelity and geometric richness across multiple benchmarks. The approach is highly efficient to train, flexible to deploy, and provides a novel low-cost pathway toward high-fidelity 3D content generation.

Existing 3D Gaussian Splatting (3DGS) methods suffer from poor hardware adaptability: lightweight variants compromise reconstruction quality, while high-fidelity approaches demand excessive GPU memory, hindering deployment across heterogeneous devices. To address this, we propose Flexible Levels of Detail (FLoD), the first scalable LoD architecture explicitly designed for 3DGS—enabling single-model, multi-granularity reconstruction and dynamic Gaussian count adjustment. Our method integrates differentiable Gaussian parameterization, adaptive density control, hierarchical scene representation, and real-time rendering optimization to achieve fine-grained trade-offs between visual quality and memory consumption. FLoD is framework-agnostic and compatible with mainstream systems including Instant-NGP and GaussianMamba. On resource-constrained devices, it reduces memory usage by up to 67% while maintaining PSNR > 28 dB. To our knowledge, this is the first work enabling high-quality real-time 3DGS rendering across diverse hardware platforms.

This work addresses the challenge of jointly preserving global structure and local geometric detail in high-fidelity 3D shape generation. We propose a two-stage diffusion framework: the first stage generates coarse-grained voxelized global structures, while the second stage refines geometric details via spatially localized voxel queries and Rotary Position Embedding (RoPE) for precise spatial anchoring. Innovatively, we introduce watertightness-preserving preprocessing and a geometry-decoupled refinement mechanism to ensure topological integrity and surface fidelity under resource constraints. The method integrates 3D diffusion modeling, voxel-based representation, RoPE-enabled spatial localization, watertight mesh repair, and multi-level data augmentation. Trained solely on public 3D datasets, our approach achieves state-of-the-art geometric quality—significantly outperforming existing open-source methods—while supporting end-to-end high-quality generation and full open-source reproducibility.

This work addresses the challenge of efficiently rendering sparse microstructured materials—such as fibrous or brushed metals—in volume rendering, which typically demand high-resolution representations yet incur substantial computational costs with conventional voxelization and multiscale rendering approaches. To overcome this, the authors propose an efficient parallel voxelization method that integrates hierarchical SGGX clustering to construct a level-of-detail (LoD) representation, significantly accelerating multiscale data aggregation and rendering. Implemented in CUDA, the approach supports both triangle meshes and explicit fiber models and is embedded within an SGGX distribution–driven LoD path tracing framework. Experimental results demonstrate that the method achieves a superior balance between rendering quality and performance across a range of microstructured materials compared to baseline techniques.

This work proposes an efficient software rasterization method for dense, opaque meshes—comprising hundreds of millions to billions of triangles—as commonly encountered in photogrammetry and related applications, without requiring prebuilt acceleration structures. The approach employs a three-stage CUDA compute shader pipeline: small triangles are processed directly in the first stage using atomicMin operations to record the nearest fragments, while large triangles are deferred to subsequent stages. Compared to Vulkan hardware rasterization, the method achieves 2–5× speedup for single-instance scenes and up to 12× acceleration with instanced rendering, substantially outperforming existing solutions, although it remains approximately an order of magnitude slower on low-polygon-count meshes.

To address the challenge of balancing geometric accuracy and visual fidelity in large-scale 3D spatial data for defense applications, this paper proposes a hierarchical hybrid representation architecture integrating classical geometric modeling with neural rendering. The architecture employs triangle meshes and voxel grids to ensure foundational geometric fidelity, while leveraging 3D Gaussian splatting and Neural Radiance Fields (NeRF) for high-fidelity photorealistic rendering at the upper layer. A unified scene management framework enables multi-granularity co-optimization across representations. Compared to purely geometric or purely neural approaches, our method achieves significant improvements: 2.1× acceleration in computational efficiency and +3.7 dB PSNR gain in rendering quality—particularly beneficial for line-of-sight analysis, physics-based simulation, and real-time visualization. It supports scalable modeling and interactive rendering of scenes with up to hundreds of millions of polygons, establishing a new paradigm for military digital twins that simultaneously delivers geometric precision, computational efficiency, and visual realism.

Existing 3D generation methods struggle to meet production-grade requirements for real-time interactive applications, such as consistent topology, UV unwrapping, physically based rendering (PBR) materials, skeletal rigging, and physically plausible scene layout. To address this gap, this work proposes a two-dimensional taxonomy centered on asset production pipelines—structured by asset type and production stage—and systematically constructs a comprehensive generation framework encompassing geometry synthesis, topology optimization, UV parameterization, PBR appearance modeling, skeletal rigging, and physics-aware scene assembly. The authors further introduce a cross-dimensional evaluation protocol to rigorously assess the direct usability of generated assets in game engines and simulation platforms. Their analysis highlights critical challenges in data quality, controllable generation, and end-to-end assetization, underscoring the pivotal role of deployable 3D content as foundational infrastructure for embodied intelligence and interactive world models.