MacDiff: Unified Skeleton Modeling with Masked Conditional Diffusion

TL;DR

MacDiff introduces a novel diffusion-based framework for skeleton modeling that improves representation learning and data augmentation, outperforming existing methods on key benchmarks.

Contribution

The paper pioneers the use of diffusion models for skeleton representation learning, integrating contrastive and generative objectives for enhanced performance.

Findings

Achieves state-of-the-art results on representation benchmarks.

Enhances fine-tuning performance with diffusion-based data augmentation.

Demonstrates the effectiveness of masked diffusion in skeleton modeling.

Abstract

Self-supervised learning has proved effective for skeleton-based human action understanding. However, previous works either rely on contrastive learning that suffers false negative problems or are based on reconstruction that learns too much unessential low-level clues, leading to limited representations for downstream tasks. Recently, great advances have been made in generative learning, which is naturally a challenging yet meaningful pretext task to model the general underlying data distributions. However, the representation learning capacity of generative models is under-explored, especially for the skeletons with spacial sparsity and temporal redundancy. To this end, we propose Masked Conditional Diffusion (MacDiff) as a unified framework for human skeleton modeling. For the first time, we leverage diffusion models as effective skeleton representation learners. Specifically, we train a diffusion decoder conditioned on the representations extracted by a semantic encoder. Random masking is applied to encoder inputs to introduce a information bottleneck and remove redundancy of skeletons. Furthermore, we theoretically demonstrate that our generative objective involves the contrastive learning objective which aligns the masked and noisy views. Meanwhile, it also enforces the representation to complement for the noisy view, leading to better generalization performance. MacDiff achieves state-of-the-art performance on representation learning benchmarks while maintaining the competence for generative tasks. Moreover, we leverage the diffusion model for data augmentation, significantly enhancing the fine-tuning performance in scenarios with scarce labeled data. Our project is available at https://lehongwu.github.io/ECCV24MacDiff/.