面向分解式人体运动生成的能量扩散模型
Towards Decompositional Human Motion Generation with Energy-Based Diffusion Models
把能量模型与扩散采样结合，用组合分布刻画多个运动概念的并集，从而在无逐概念真值的情况下把整体运动拆成可复用基元；这套「能量式组合分布 + 分解训练」的思路可直接迁移到视觉中的组合式视频/动作生成与概念解耦。
原文依据 we propose DeMoGen, a compositional training paradigm for decompositional learning that employs an energy-based diffusion model
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论人体运动具有组合性：复杂行为可以被描述为更简单基元的组合。然而，现有方法主要关注前向建模，例如学习从文本到运动的整体映射，或由一组运动概念组合出复杂运动。本文考虑相反的视角：将一个整体运动分解为语义上有意义的子成分。我们提出 DeMoGen，一种采用能量扩散模型的组合式训练范式，用于分解式学习。该能量形式直接刻画多个运动概念的组合分布，使模型无需依赖各个概念的真值运动即可发现这些概念。在该范式下，我们引入三种训练变体以促进对运动的分解式理解：1. DeMoGen-Exp 在分解后的文本提示上显式训练；2. DeMoGen-OSS 执行正交自监督分解；3. DeMoGen-SC 强制原始文本嵌入与分解文本嵌入之间的语义一致性。这些变体使我们的方法能够从复杂运动序列中解耦出可复用的运动基元。我们还证明，分解得到的运动概念可以被灵活重组，以生成多样且新颖的运动，并泛化到训练分布之外。此外，我们构建了一个文本分解数据集以支持组合式训练，作为促进文本到运动生成与运动组合的扩展资源。
英文原文摘要 1536 字符
Human motions are compositional: complex behaviors can be described as combinations of simpler primitives. However, existing approaches primarily focus on forward modeling, e.g., learning holistic mappings from text to motion or composing a complex motion from a set of motion concepts. In this paper, we consider the inverse perspective: decomposing a holistic motion into semantically meaningful sub-components. We propose DeMoGen, a compositional training paradigm for decompositional learning that employs an energy-based diffusion model. This energy formulation directly captures the composed distribution of multiple motion concepts, enabling the model to discover them without relying on ground-truth motions for individual concepts. Within this paradigm, we introduce three training variants to encourage a decompositional understanding of motion: 1. DeMoGen-Exp explicitly trains on decomposed text prompts; 2. DeMoGen-OSS performs orthogonal self-supervised decomposition; 3. DeMoGen-SC enforces semantic consistency between original and decomposed text embeddings. These variants enable our approach to disentangle reusable motion primitives from complex motion sequences. We also demonstrate that the decomposed motion concepts can be flexibly recombined to generate diverse and novel motions, generalizing beyond the training distribution. Additionally, we construct a text-decomposed dataset to support compositional training, serving as an extended resource to facilitate text-to-motion generation and motion composition.