从单张图像零样本重建具有布料动力学的可动画 3D 化身
Zero-Shot Reconstruction of Animatable 3D Avatars with Cloth Dynamics from a Single Image
提出基于 Transformer 的前馈架构直接预测动态 3D 高斯形变，结合静态到动态的知识迁移与 LoRA 微调，为单图人体化身重建提供可复用的零样本方案；DynaFlow 损失利用光流引导几何线索，可迁移至其他动态重建任务。
原文依据 DynaAvatar employs a Transformer-based feed-forward architecture that directly predicts dynamic 3D Gaussian deformations without subject-specific optimization.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论现有的单图像 3D 人体化身方法主要依赖刚性关节变换，限制了其建模真实布料动力学的能力。我们提出 DynaAvatar，一个零样本框架，能够从单张图像重建具有运动相关布料动力学的可动画 3D 人体化身。DynaAvatar 在大规模多人运动数据集上训练，采用基于 Transformer 的前馈架构，直接预测动态 3D 高斯形变，无需针对特定对象进行优化。为克服动态捕捉数据的稀缺性，我们引入静态到动态的知识迁移策略：在大规模静态捕捉数据上预训练的 Transformer 提供强大的几何与外观先验，通过轻量级低秩适配在动态捕捉数据上微调，高效地适应运动相关形变。我们进一步提出 DynaFlow 损失，一种光流引导的目标函数，在渲染空间中为布料动力学提供可靠的运动方向几何线索。最后，我们重新标注了现有动态捕捉数据集中缺失或噪声的 SMPL-X 拟合，因为大多数公开动态捕捉数据集包含不完整或不可靠的拟合，不适合训练高质量 3D 化身重建模型。实验表明，DynaAvatar 生成视觉丰富且可泛化的动画，优于先前方法。代码、预训练模型和重新标注将发布。
英文原文摘要 1452 字符
Existing single-image 3D human avatar methods primarily rely on rigid joint transformations, limiting their ability to model realistic cloth dynamics. We present DynaAvatar, a zero-shot framework that reconstructs animatable 3D human avatars with motion-dependent cloth dynamics from a single image. Trained on large-scale multi-person motion datasets, DynaAvatar employs a Transformer-based feed-forward architecture that directly predicts dynamic 3D Gaussian deformations without subject-specific optimization. To overcome the scarcity of dynamic captures, we introduce a static-to-dynamic knowledge transfer strategy: a Transformer pretrained on large-scale static captures provides strong geometric and appearance priors, which are efficiently adapted to motion-dependent deformations through lightweight LoRA fine-tuning on dynamic captures. We further propose the DynaFlow loss, an optical flow-guided objective that provides reliable motion-direction geometric cues for cloth dynamics in rendered space. Finally, we reannotate the missing or noisy SMPL-X fittings in existing dynamic capture datasets, as most public dynamic capture datasets contain incomplete or unreliable fittings that are unsuitable for training high-quality 3D avatar reconstruction models. Experiments demonstrate that DynaAvatar produces visually rich and generalizable animations, outperforming prior methods. Code, pretrained models, and reannotations will be released.