用分块稀疏网格实现可扩展3D建模，内存更低、推理更快。
Beyond Global Latents: Chunk-Based Sparse Grid VAE for Scalable 3D Modeling
- 按局部块组织编码器和解码器，灵活处理不同尺寸输入输出。
- 在512³到1536³分辨率下性能优于或媲美主流方法，峰值内存降低。
- 适合需要高精度3D重建且追求高效推理的场景。
稀疏体素网格能保留详细3D重建所需的结构，但其内存随分辨率快速上升，因活跃表面单元增多。我们提出ChunkVAE，一种基于局部块而非全局潜在体积的稀疏网格变分自编码器。局部学习算子允许独立选择编码器与解码器的分区，使推理块大小可不同于训练。两种互补数据操作使这种灵活性可行：平衡二叉物体分割分布活跃单元并限制重复重叠；S-曲线加权拼接在组装全局潜在表示或重建时衰减不可靠边界特征。在三个物体基准上，ChunkVAE在512³至1536³分辨率范围内表现与强基线相当或更优；更小的块降低峰值内存占用并缩短单块计算时间，支持更快并行推理。稳定拼接的潜在表示和提升的图像到3D指标表明，局部压缩可在保持下游所需全局接口的同时实现几何可扩展性。
原文摘要 · Abstract (English)
Sparse voxel grids preserve the spatial structure needed for detailed 3D reconstruction, but their memory still grows rapidly with resolution as active surface cells increase. We introduce ChunkVAE, a sparse grid variational autoencoder organized around local chunks rather than a global latent volume. Local learned operators permit independently chosen encoder and decoder partitions and allow inference chunk sizes to differ from training. Two complementary data operators make this flexibility practical: Balanced Binary Object Partitioning distributes active cells while limiting replicated overlap, while S-Curve weighted stitching attenuates unreliable boundary features when assembling a global latent or reconstruction. Across three object benchmarks, ChunkVAE is competitive with or better than strong baselines from $512^3$ to $1536^3$; smaller chunks lower peak allocated memory and shorten per-chunk compute, enabling faster parallel inference. Stable stitched latents and improved image to 3D metrics indicate that local compression can scale geometry while retaining the global interface required downstream.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。