用于头部化身创建与编辑的前馈高斯配准
Feed-forward Gaussian Registration for Head Avatar Creation and Editing
用 transformer 在模板网格固定 UV 布局下前馈预测高斯泼溅纹理，并通过配准引导注意力让每个 UV 词元只关注其对应网格区域的图像词元，从而在 0.5 秒/帧内建立跨帧与跨主体的稠密对应；这套「UV 空间 + 对应性注意力」可直接复用到任何需要多视角一致纹理与免优化跟踪的人体/人脸建模流程。
原文依据 each UV map token attends exclusively to image tokens depicting its corresponding mesh region
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 MATCH（Multi-view Avatars from Topologically Corresponding Heads），一种用于高质量头部化身创建与编辑的多视角高斯配准方法。最优的多视角头部化身需要耗时的头部跟踪，随后还要进行昂贵的化身优化，总创建时间往往超过一天。MATCH 则直接从标定好的多视角图像中预测处于对应关系下的高斯泼溅纹理，每帧仅需 0.5 秒。跨帧学习到的同一主体内部对应关系使我们能够快速构建个性化头部化身，而跨主体的对应关系则支持表情迁移、免优化跟踪、语义编辑和身份插值等多种应用。我们以端到端方式学习建立此类对应关系，所用模型基于 transformer，在模板网格固定的 UV 布局中预测高斯泼溅的纹理。为此，我们引入了一种新颖的配准引导注意力模块，其中每个 UV 图词元只关注描绘其对应网格区域的图像词元。MATCH 在新视角合成、几何配准和头部化身生成上均优于现有方法，其中最后一项比质量最接近的基线快 10 倍。代码与模型权重见 https://malteprinzler.github.io/projects/match。
英文原文摘要 1395 字符
We present MATCH (Multi-view Avatars from Topologically Corresponding Heads), a multi-view Gaussian registration method for high-quality head avatar creation and editing. State-of-the-art multi-view head avatars require time-consuming head tracking, which is followed by an expensive avatar optimization, often resulting in a total creation time that exceeds one day. MATCH instead directly predicts Gaussian splat textures in correspondence from calibrated multi-view images in 0.5 seconds per frame. While the learned intra-subject correspondence across frames allows us to quickly build personalized head avatars, correspondence across subjects enables various applications such as expression transfer, optimization-free tracking, semantic editing, and identity interpolation. We learn to establish such correspondences end-to-end, with a transformer-based model that predicts textures of Gaussian splats in the fixed UV layout of a template mesh. To this end, we introduce a novel registration-guided attention block, in which each UV map token attends exclusively to image tokens depicting its corresponding mesh region. MATCH outperforms existing methods for novel-view synthesis, geometry registration, and head avatar generation, the latter being 10xfaster than the qualitatively closest baseline. Code and model weights are available under https://malteprinzler.github.io/projects/match