无需相机参数，从稀疏视角重建高质量3D高斯点云
StructSplat: Generalizable 3D Gaussian Splatting from Uncalibrated Sparse Views
- 分离几何、语义与纹理，分步建模提升精度
- 在DL3DV上达28.045 PSNR，优于现有方法5.67 dB
- 支持跨数据集泛化，适合真实场景三维重建
我们提出StructSplat，一种直接处理未标定图像的前馈式通用3D高斯重建框架，无需依赖相机参数。现有方法通常需逐场景优化或假设已知相机位姿，且常将几何与外观混合建模，限制重建质量与泛化能力。我们的核心思想是采用结构化表示，明确区分几何、语义和纹理在重建中的角色。具体包括：引入像素对齐特征注入机制，实现从2D观测中精准建模纹理；融入语义感知先验，增强全局一致性；设计相机对齐策略，防止信息泄露，提升泛化性。实验表明，该方法在挑战性基准上显著超越现有方法：在DL3DV上达到28.045 PSNR，较AnySplat（22.377）提升+5.67 dB；跨数据集评估中，在ACID上领先+1.94 dB，RealEstate10K上领先+1.72 dB。
原文摘要 · Abstract (English)
We present StructSplat, a feed-forward and generalizable 3D Gaussian reconstruction framework that operates directly on uncalibrated images without requiring camera parameters. Existing methods either rely on per-scene optimization or assume known camera poses, and often entangle geometry and appearance within a unified backbone, limiting reconstruction fidelity and generalization. Our key idea is to adopt a structured representation that organizes geometry, semantic, and texture cues with explicit roles in the reconstruction process. Specifically, we introduce a pixel-aligned feature injection mechanism to enable accurate texture modeling from 2D observations, incorporate semantic-aware priors to improve global consistency, and design a camera alignment strategy to prevent information leakage and improve generalization. Experiments show that our method significantly outperforms prior approaches on challenging benchmarks. On DL3DV, our method achieves 28.045 PSNR, surpassing AnySplat (22.377) by +5.67 dB. In cross-dataset evaluation, our method achieves +1.94 dB over AnySplat on ACID and +1.72 dB on RealEstate10K. Project page: https://structsplat.github.io Code: https://github.com/J-C-Zhao/StructSplat
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。