DIVE: Inverting Conditional Diffusion Models for Discriminative Tasks

doi:10.1109/TMM.2025.3623508

CSpace

	DIVE: Inverting Conditional Diffusion Models for Discriminative Tasks
	Li, Yinqi 1,2; Chang, Hong 1,2; Hou, Ruibing 1; Shan, Shiguang 1,2; Chen, Xilin 1,2
	2026
发表期刊	IEEE TRANSACTIONS ON MULTIMEDIA
ISSN	1520-9210
卷号	28 页码:297-308
摘要	Diffusion models have shown remarkable progress in various generative tasks such as image and video generation. This paper studies the problem of leveraging pretrained diffusion models for performing discriminative tasks. Specifically, we extend the discriminative capability of pretrained frozen generative diffusion models from the classification task (Li et al., 2023), (Clark et al., 2023) to the more complex object detection task, by "inverting" a pretrained layout-to-image diffusion model. To this end, a gradient-based discrete optimization approach for replacing the heavy prediction enumeration process, and a prior distribution model for making more accurate use of the Bayes' rule, are proposed respectively. Empirical results show that this method is on par with basic discriminative object detection baselines on COCO dataset. In addition, our method can greatly speed up the previous diffusion-based method (Li et al., 2023), (Clark et al., 2023) for classification without sacrificing accuracy. Code and models are available at https://github.com/LiYinqi/DIVE.
关键词	Diffusion models Object detection Layout Noise Training Noise reduction Vocabulary Dogs Bicycles Image synthesis Diffusion model generative modeling discriminative Task object detection visual recognition
DOI	10.1109/TMM.2025.3623508
收录类别	SCI
语种	英语
WOS研究方向	Computer Science ; Telecommunications
WOS类目	Computer Science, Information Systems ; Computer Science, Software Engineering ; Telecommunications
WOS记录号	WOS:001658631000033
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/42842
专题	中国科学院计算技术研究所
通讯作者	Chang, Hong
作者单位	1.Chinese Acad Sci, Inst Comp Technol, Key Lab Intelligent Informat Proc, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100049, Peoples R China
推荐引用方式 GB/T 7714	Li, Yinqi,Chang, Hong,Hou, Ruibing,et al. DIVE: Inverting Conditional Diffusion Models for Discriminative Tasks[J]. IEEE TRANSACTIONS ON MULTIMEDIA,2026,28:297-308.
APA	Li, Yinqi,Chang, Hong,Hou, Ruibing,Shan, Shiguang,&Chen, Xilin.(2026).DIVE: Inverting Conditional Diffusion Models for Discriminative Tasks.IEEE TRANSACTIONS ON MULTIMEDIA,28,297-308.
MLA	Li, Yinqi,et al."DIVE: Inverting Conditional Diffusion Models for Discriminative Tasks".IEEE TRANSACTIONS ON MULTIMEDIA 28(2026):297-308.