MoRE：三维视觉几何重建遇上混合专家
MoRE: 3D Visual Geometry Reconstruction Meets Mixture-of-Experts
用 MoE 动态路由把深度、法线等几何任务分派给专家，并配一个基于置信度的深度细化模块，可作为构建可扩展稠密几何基础模型、在多任务间共享骨干网络时的架构与损失设计参考。
原文依据 a dense 3D visual foundation model based on a Mixture-of-Experts (MoE) architecture that dynamically routes features to task-specific experts
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期语言与视觉领域的进展表明，扩大模型容量能够持续提升各类任务的性能。在三维视觉几何重建中，大规模训练同样被证明有助于学习通用的表征。然而，由于几何监督的复杂性以及三维数据的多样性，进一步扩展三维模型规模颇具挑战。为克服这些限制，我们提出 MoRE，一个基于混合专家（MoE）架构的稠密三维视觉基础模型，它将特征动态路由至任务特定的专家，使各专家能够专注于互补的数据方面，从而同时提升可扩展性与适应性。为提升真实场景下的鲁棒性，MoRE 引入基于置信度的深度细化模块，以稳定并精化几何估计。此外，它将稠密语义特征与全局对齐的三维骨干网络表征相结合，以实现高保真的表面法线预测。MoRE 进一步通过定制化的损失函数进行优化，以确保在多样输入与多种几何任务上实现稳健学习。大量实验表明，MoRE 在多个基准上取得最优性能，并能在不引入额外计算的情况下有效支撑下游应用。
英文原文摘要 1323 字符
Recent advances in language and vision have demonstrated that scaling up model capacity consistently improves performance across diverse tasks.In 3D visual geometry reconstruction, large-scale training has likewise proven effective for learning versatile representations.However, further scaling of 3D models is challenging due to the complexity of geometric supervision and the diversity of 3D data. To overcome these limitations, we propose MoRE, a dense 3D visual foundation model based on a Mixture-of-Experts (MoE) architecture that dynamically routes features to task-specific experts, allowing them to specialize in complementary data aspects and enhance both scalability and adaptability.Aiming to improve robustness under real-world conditions, MoRE incorporates a confidence-based depth refinement module that stabilizes and refines geometric estimation.In addition, it integrates dense semantic features with globally aligned 3D backbone representations for high-fidelity surface normal prediction.MoRE is further optimized with tailored loss functions to ensure robust learning across diverse inputs and multiple geometric tasks.Extensive experiments demonstrate that MoRE achieves state-of-the-art performance across multiple benchmarks and supports effective downstream applications without extra computation.