OMG-Avatar：单样本多细节层次高斯头部化身
OMG-Avatar: One-shot Multi-LOD Gaussian Head Avatar
用深度缓冲引导的 Transformer 全局特征与投影采样局部特征融合，在单张图像上 0.2 秒内重建可动画高斯头部，并支持多级细节（LOD）输出；其多区域分解与跨区域组合思路可直接迁移到人体/肩部等非 3DMM 区域的单样本重建。
原文依据 a novel One-shot method that leverages a Multi-LOD (Level-of-Detail) Gaussian representation for animatable 3D head reconstruction from a single image in 0.2s
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论我们提出 OMG-Avatar，一种新颖的单样本方法，利用多细节层次（Level-of-Detail，LOD）高斯表示，在 0.2 秒内从单张图像重建可动画的三维头部。本方法使用统一模型实现 LOD 头部化身建模，可适配多样的硬件能力与推理速度需求。为同时捕捉全局与局部面部特征，我们采用基于 Transformer 的架构进行全局特征提取，并使用基于投影的采样获取局部特征。这些特征在深度缓冲的引导下有效融合，从而保证遮挡的合理性。我们进一步引入由粗到细的学习范式以支持细节层次功能，并增强对层次化细节的感知。针对 3DMM 在建模肩部等非头部区域方面的局限，我们引入多区域分解方案，将头部与肩部分别预测，再通过跨区域组合进行整合。大量实验表明，OMG-Avatar 在重建质量、重演性能与计算效率上均优于最优方法。
英文原文摘要 1195 字符
We propose OMG-Avatar, a novel One-shot method that leverages a Multi-LOD (Level-of-Detail) Gaussian representation for animatable 3D head reconstruction from a single image in 0.2s. Our method enables LOD head avatar modeling using a unified model that accommodates diverse hardware capabilities and inference speed requirements. To capture both global and local facial characteristics, we employ a transformer-based architecture for global feature extraction and projection-based sampling for local feature acquisition. These features are effectively fused under the guidance of a depth buffer, ensuring occlusion plausibility. We further introduce a coarse-to-fine learning paradigm to support Level-of-Detail functionality and enhance the perception of hierarchical details. To address the limitations of 3DMMs in modeling non-head regions such as the shoulders, we introduce a multi-region decomposition scheme in which the head and shoulders are predicted separately and then integrated through cross-region combination. Extensive experiments demonstrate that OMG-Avatar outperforms state-of-the-art methods in reconstruction quality, reenactment performance, and computational efficiency.