Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Beg, Hajra Anwar, Chopin, Baptiste, Tang, Hao, Daoudi, Mohamed
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2512.00208
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915759432663040
author Beg, Hajra Anwar
Chopin, Baptiste
Tang, Hao
Daoudi, Mohamed
author_facet Beg, Hajra Anwar
Chopin, Baptiste
Tang, Hao
Daoudi, Mohamed
contents We present ReactionMamba, a novel framework for generating long 3D human reaction motions. Reaction-Mamba integrates a motion VAE for efficient motion encoding with Mamba-based state-space models to decode temporally consistent reactions. This design enables ReactionMamba to generate both short sequences of simple motions and long sequences of complex motions, such as dance and martial arts. We evaluate ReactionMamba on three datasets--NTU120-AS, Lindy Hop, and InterX--and demonstrate competitive performance in terms of realism, diversity, and long-sequence generation compared to previous methods, including InterFormer, ReMoS, and Ready-to-React, while achieving substantial improvements in inference speed.
format Preprint
id arxiv_https___arxiv_org_abs_2512_00208
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ReactionMamba: Generating Short & Long Human Reaction Sequences
Beg, Hajra Anwar
Chopin, Baptiste
Tang, Hao
Daoudi, Mohamed
Computer Vision and Pattern Recognition
We present ReactionMamba, a novel framework for generating long 3D human reaction motions. Reaction-Mamba integrates a motion VAE for efficient motion encoding with Mamba-based state-space models to decode temporally consistent reactions. This design enables ReactionMamba to generate both short sequences of simple motions and long sequences of complex motions, such as dance and martial arts. We evaluate ReactionMamba on three datasets--NTU120-AS, Lindy Hop, and InterX--and demonstrate competitive performance in terms of realism, diversity, and long-sequence generation compared to previous methods, including InterFormer, ReMoS, and Ready-to-React, while achieving substantial improvements in inference speed.
title ReactionMamba: Generating Short & Long Human Reaction Sequences
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.00208