Saved in:
Bibliographic Details
Main Authors: Tanke, Julian, Shibuya, Takashi, Uchida, Kengo, Saito, Koichi, Mitsufuji, Yuki
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.09827
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915287921590272
author Tanke, Julian
Shibuya, Takashi
Uchida, Kengo
Saito, Koichi
Mitsufuji, Yuki
author_facet Tanke, Julian
Shibuya, Takashi
Uchida, Kengo
Saito, Koichi
Mitsufuji, Yuki
contents Generating realistic dyadic human motion from text descriptions presents significant challenges, particularly for extended interactions that exceed typical training sequence lengths. While recent transformer-based approaches have shown promising results for short-term dyadic motion synthesis, they struggle with longer sequences due to inherent limitations in positional encoding schemes. In this paper, we introduce Dyadic Mamba, a novel approach that leverages State-Space Models (SSMs) to generate high-quality dyadic human motion of arbitrary length. Our method employs a simple yet effective architecture that facilitates information flow between individual motion sequences through concatenation, eliminating the need for complex cross-attention mechanisms. We demonstrate that Dyadic Mamba achieves competitive performance on standard short-term benchmarks while significantly outperforming transformer-based approaches on longer sequences. Additionally, we propose a new benchmark for evaluating long-term motion synthesis quality, providing a standardized framework for future research. Our results demonstrate that SSM-based architectures offer a promising direction for addressing the challenging task of long-term dyadic human motion synthesis from text descriptions.
format Preprint
id arxiv_https___arxiv_org_abs_2505_09827
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Dyadic Mamba: Long-term Dyadic Human Motion Synthesis
Tanke, Julian
Shibuya, Takashi
Uchida, Kengo
Saito, Koichi
Mitsufuji, Yuki
Computer Vision and Pattern Recognition
Generating realistic dyadic human motion from text descriptions presents significant challenges, particularly for extended interactions that exceed typical training sequence lengths. While recent transformer-based approaches have shown promising results for short-term dyadic motion synthesis, they struggle with longer sequences due to inherent limitations in positional encoding schemes. In this paper, we introduce Dyadic Mamba, a novel approach that leverages State-Space Models (SSMs) to generate high-quality dyadic human motion of arbitrary length. Our method employs a simple yet effective architecture that facilitates information flow between individual motion sequences through concatenation, eliminating the need for complex cross-attention mechanisms. We demonstrate that Dyadic Mamba achieves competitive performance on standard short-term benchmarks while significantly outperforming transformer-based approaches on longer sequences. Additionally, we propose a new benchmark for evaluating long-term motion synthesis quality, providing a standardized framework for future research. Our results demonstrate that SSM-based architectures offer a promising direction for addressing the challenging task of long-term dyadic human motion synthesis from text descriptions.
title Dyadic Mamba: Long-term Dyadic Human Motion Synthesis
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.09827