Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Xiao, Chen, Qi, Peng, Xiulian, Yu, Kai, Chen, Xie, Lu, Yan
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2509.08376
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918138700890112
author Li, Xiao
Chen, Qi
Peng, Xiulian
Yu, Kai
Chen, Xie
Lu, Yan
author_facet Li, Xiao
Chen, Qi
Peng, Xiulian
Yu, Kai
Chen, Xie
Lu, Yan
contents We propose a novel and general framework to disentangle video data into its dynamic motion and static content components. Our proposed method is a self-supervised pipeline with less assumptions and inductive biases than previous works: it utilizes a transformer-based architecture to jointly generate flexible implicit features for frame-wise motion and clip-wise content, and incorporates a low-bitrate vector quantization as an information bottleneck to promote disentanglement and form a meaningful discrete motion space. The bitrate-controlled latent motion and content are used as conditional inputs to a denoising diffusion model to facilitate self-supervised representation learning. We validate our disentangled representation learning framework on real-world talking head videos with motion transfer and auto-regressive motion generation tasks. Furthermore, we also show that our method can generalize to other types of video data, such as pixel sprites of 2D cartoon characters. Our work presents a new perspective on self-supervised learning of disentangled video representations, contributing to the broader field of video analysis and generation.
format Preprint
id arxiv_https___arxiv_org_abs_2509_08376
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Bitrate-Controlled Diffusion for Disentangling Motion and Content in Video
Li, Xiao
Chen, Qi
Peng, Xiulian
Yu, Kai
Chen, Xie
Lu, Yan
Computer Vision and Pattern Recognition
We propose a novel and general framework to disentangle video data into its dynamic motion and static content components. Our proposed method is a self-supervised pipeline with less assumptions and inductive biases than previous works: it utilizes a transformer-based architecture to jointly generate flexible implicit features for frame-wise motion and clip-wise content, and incorporates a low-bitrate vector quantization as an information bottleneck to promote disentanglement and form a meaningful discrete motion space. The bitrate-controlled latent motion and content are used as conditional inputs to a denoising diffusion model to facilitate self-supervised representation learning. We validate our disentangled representation learning framework on real-world talking head videos with motion transfer and auto-regressive motion generation tasks. Furthermore, we also show that our method can generalize to other types of video data, such as pixel sprites of 2D cartoon characters. Our work presents a new perspective on self-supervised learning of disentangled video representations, contributing to the broader field of video analysis and generation.
title Bitrate-Controlled Diffusion for Disentangling Motion and Content in Video
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.08376