Saved in:
Bibliographic Details
Main Authors: Chen, Changyou, Ding, Han, Sisman, Bunyamin, Xu, Yi, Xie, Ouye, Yao, Benjamin Z., Tran, Son Dinh, Zeng, Belinda
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2407.17571
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912044958089216
author Chen, Changyou
Ding, Han
Sisman, Bunyamin
Xu, Yi
Xie, Ouye
Yao, Benjamin Z.
Tran, Son Dinh
Zeng, Belinda
author_facet Chen, Changyou
Ding, Han
Sisman, Bunyamin
Xu, Yi
Xie, Ouye
Yao, Benjamin Z.
Tran, Son Dinh
Zeng, Belinda
contents Diffusion-based generative modeling has been achieving state-of-the-art results on various generation tasks. Most diffusion models, however, are limited to a single-generation modeling. Can we generalize diffusion models with the ability of multi-modal generative training for more generalizable modeling? In this paper, we propose a principled way to define a diffusion model by constructing a unified multi-modal diffusion model in a common diffusion space. We define the forward diffusion process to be driven by an information aggregation from multiple types of task-data, e.g., images for a generation task and labels for a classification task. In the reverse process, we enforce information sharing by parameterizing a shared backbone denoising network with additional modality-specific decoder heads. Such a structure can simultaneously learn to generate different types of multi-modal data with a multi-task loss, which is derived from a new multi-modal variational lower bound that generalizes the standard diffusion model. We propose several multimodal generation settings to verify our framework, including image transition, masked-image training, joint image-label and joint image-representation generative modeling. Extensive experimental results on ImageNet indicate the effectiveness of our framework for various multi-modal generative modeling, which we believe is an important research direction worthy of more future explorations.
format Preprint
id arxiv_https___arxiv_org_abs_2407_17571
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Diffusion Models For Multi-Modal Generative Modeling
Chen, Changyou
Ding, Han
Sisman, Bunyamin
Xu, Yi
Xie, Ouye
Yao, Benjamin Z.
Tran, Son Dinh
Zeng, Belinda
Computer Vision and Pattern Recognition
Diffusion-based generative modeling has been achieving state-of-the-art results on various generation tasks. Most diffusion models, however, are limited to a single-generation modeling. Can we generalize diffusion models with the ability of multi-modal generative training for more generalizable modeling? In this paper, we propose a principled way to define a diffusion model by constructing a unified multi-modal diffusion model in a common diffusion space. We define the forward diffusion process to be driven by an information aggregation from multiple types of task-data, e.g., images for a generation task and labels for a classification task. In the reverse process, we enforce information sharing by parameterizing a shared backbone denoising network with additional modality-specific decoder heads. Such a structure can simultaneously learn to generate different types of multi-modal data with a multi-task loss, which is derived from a new multi-modal variational lower bound that generalizes the standard diffusion model. We propose several multimodal generation settings to verify our framework, including image transition, masked-image training, joint image-label and joint image-representation generative modeling. Extensive experimental results on ImageNet indicate the effectiveness of our framework for various multi-modal generative modeling, which we believe is an important research direction worthy of more future explorations.
title Diffusion Models For Multi-Modal Generative Modeling
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2407.17571