Jiehua Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jiehua Zhang has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Chenggang Yan
Hangzhou Dianzi University, Hangzhou, China
14 shared publications - Liang Li
Chinese Academy of Sciences
14 shared publications - Yaoqi Sun
Lishui University, Lishui, China
5 shared publications - Xichun Sheng
Macao Polytechnic University, Macao, China
4 shared publications - Yihong Gong
Xi'an Jiaotong University, Xi'an, China
4 shared publications - Wei Ke
School of Software Engineering, Xian Jiaotong University, Xian, China
3 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 9687-9696
Fine-Grained Visual and Text Forgery Distillation for Media Manipulation Detection
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 9687-9696
Open AccessFine-Grained Visual and Text Forgery Distillation for Media Manipulation Detection
- 27 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. PP (15209210) , 1-11
Semantic Distribution and Authenticity Discrepancy Alignment for AI-Generated Image Detection
Generative models have achieved remarkable success in producing vivid images. Compared with real images, generated ones still show different semantic structures that features with different semantic classes collapse as a single cluster. Pioneer works leverage the discrepancy of semantic structure in fixed high-level semantic feature space to identify forgery images. Nevertheless, such frozen pre-trained representation models are insensitive to subtle forgery traces. Meanwhile, vanilla fine-tuning methods can distort the pre-trained semantic knowledge and collapse to the real-fake binary distribution, losing generalization capability in newly emerged generative models. In this paper, we propose the semantic distribution and authenticity discrepancy alignment algorithm (STERM), which learns high-level semantic structures of real-world categories and low-level forgery traces for detecting AI-generated images from unseen generative models and frameworks. Specifically, we first capture semantic features of images by the frozen CLIP and further extract forgery features by a forgery encoder. Then, we propose semantic distribution alignment (SDA) to align the semantic structure of real-world categories by enforcing forgery feature distribution shifting towards the semantic feature space. Next, we introduce authenticity discrepancy alignment (ADA) to minimize the authenticity discrepancy between forgery and semantic features, constraining forgery features from collapsing into the source domain-biased distribution and learning the semantic structure of real-world categories. Extensive experiments on GAN-based and diffusion model-based datasets demonstrate the generalization capability of the proposed method. The source code is publicly available at https://github.com/freshjh/STERM.
- 27 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. PP (15209210) , 1-11
Semantic Distribution and Authenticity Discrepancy Alignment for AI-Generated Image Detection
Generative models have achieved remarkable success in producing vivid images. Compared with real images, generated ones still show different semantic structures that features with different semantic classes collapse as a single cluster. Pioneer works leverage the discrepancy of semantic structure in fixed high-level semantic feature space to identify forgery images. Nevertheless, such frozen pre-trained representation models are insensitive to subtle forgery traces. Meanwhile, vanilla fine-tuning methods can distort the pre-trained semantic knowledge and collapse to the real-fake binary distribution, losing generalization capability in newly emerged generative models. In this paper, we propose the semantic distribution and authenticity discrepancy alignment algorithm (STERM), which learns high-level semantic structures of real-world categories and low-level forgery traces for detecting AI-generated images from unseen generative models and frameworks. Specifically, we first capture semantic features of images by the frozen CLIP and further extract forgery features by a forgery encoder. Then, we propose semantic distribution alignment (SDA) to align the semantic structure of real-world categories by enforcing forgery feature distribution shifting towards the semantic feature space. Next, we introduce authenticity discrepancy alignment (ADA) to minimize the authenticity discrepancy between forgery and semantic features, constraining forgery features from collapsing into the source domain-biased distribution and learning the semantic structure of real-world categories. Extensive experiments on GAN-based and diffusion model-based datasets demonstrate the generalization capability of the proposed method. The source code is publicly available at https://github.com/freshjh/STERM.
- 20 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28 (15209210) , 3986-4001
Hybrid Debiasing Transformer With Adaptive Regularization for Video Moment Localization
Video Moment Localization (VML) is a task that seeks to pinpoint the most pertinent segment within an untrimmed video using a linguistic query. Previous works expose the severe data bias issues in VML and note that models avoid understanding visual-textual content by adapting the timestamp distribution. The work investigates data biases from both intrinsic and extrinsic perspectives: The former arises primarily from moment boundary ambiguity and the inputoutput information imbalance. The latter is attributed to the longtail distribution and the semantic bias resulting from the limited tail samples. To reduce the issues, we develop a hybrid multimodal debiasing network with a temporal consistency constraint for VML. Firstly, we propose a multi-temporal Transformer to alleviate boundary ambiguity by merging frame-wise features into segment-wise representations and dynamically aligning with moment boundaries. Subsequently, we implement a temporal consistency constraint to accentuate action information from complex moment context and mitigate the intrinsic bias caused by information imbalance. Moreover, we develop a hybrid linguistic activation module to mitigate the long-tail bias, which offers prior guidance to emphasize distinguishing clues from tail samples. Additionally, we introduce the prior-guided Transformer to alleviate the semantic bias by learning the global semantics of sentences, thereby circumventing the tail-sample overfitting issue. Comprehensive experiments demonstrate the efficacy of our proposed method across three datasets.
- 20 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28 (15209210) , 3986-4001
Hybrid Debiasing Transformer With Adaptive Regularization for Video Moment Localization
Video Moment Localization (VML) is a task that seeks to pinpoint the most pertinent segment within an untrimmed video using a linguistic query. Previous works expose the severe data bias issues in VML and note that models avoid understanding visual-textual content by adapting the timestamp distribution. The work investigates data biases from both intrinsic and extrinsic perspectives: The former arises primarily from moment boundary ambiguity and the inputoutput information imbalance. The latter is attributed to the longtail distribution and the semantic bias resulting from the limited tail samples. To reduce the issues, we develop a hybrid multimodal debiasing network with a temporal consistency constraint for VML. Firstly, we propose a multi-temporal Transformer to alleviate boundary ambiguity by merging frame-wise features into segment-wise representations and dynamically aligning with moment boundaries. Subsequently, we implement a temporal consistency constraint to accentuate action information from complex moment context and mitigate the intrinsic bias caused by information imbalance. Moreover, we develop a hybrid linguistic activation module to mitigate the long-tail bias, which offers prior guidance to emphasize distinguishing clues from tail samples. Additionally, we introduce the prior-guided Transformer to alleviate the semantic bias by learning the global semantics of sentences, thereby circumventing the tail-sample overfitting issue. Comprehensive experiments demonstrate the efficacy of our proposed method across three datasets.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 13843-13854
Error-Guided Artifact Spotting and High-Frequency Detail Learning for AI-Generated Image Detection
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 13843-13854
Error-Guided Artifact Spotting and High-Frequency Detail Learning for AI-Generated Image Detection
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 14635-14646
Scale-Aware Domain-Shared Depth Distribution Adapter for Incremental Monocular Depth Estimation
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28, 14635-14646
Scale-Aware Domain-Shared Depth Distribution Adapter for Incremental Monocular Depth Estimation
- 3 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28 (15209210) , 1457-1468
Prompt Learning With Knowledge Regularization for Pre-Trained Vision-Language Models
Prompt learning is an effective way to adapt pre-trained models to downstream tasks by training a small number of additional learnable prompts. Recent studies address several early challenges by combining generalized knowledge from frozen pre-trained VL models with task-specific knowledge from training data as guidance for prompt learning. However, existing methods still struggle with the generalization-adaptation (GA) trade-off dilemma: excessive reliance on generalized knowledge hinders adaptation to downstream tasks, while overemphasis on task-specific knowledge undermines the inherent generalization capabilities of pre-trained models. To address this issue, we propose a novel prompt learning method called Prompt Learning with Knowledge Regularization (PLKR). PLKR effectively mitigates the GA trade-off dilemma by offering greater flexibility in adapting to task-specific knowledge while minimizing the disruption of pre-trained knowledge. Specifically, we propose category-invariant and topology-invariant knowledge regularization to preserve generalized knowledge: the former enhances category-level discriminative capabilities while allowing flexible task-specific learning, and the latter maintains global topological stability during adaptation to new tasks. Through the proposed regularization, PLKR improves the performance on both base and new tasks. We evaluate the effectiveness of our approach on four representative tasks over 11 datasets. Experimental results show our method outperforms existing SOTA methods by a large margin.
- 3 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28 (15209210) , 1457-1468
Prompt Learning With Knowledge Regularization for Pre-Trained Vision-Language Models
Prompt learning is an effective way to adapt pre-trained models to downstream tasks by training a small number of additional learnable prompts. Recent studies address several early challenges by combining generalized knowledge from frozen pre-trained VL models with task-specific knowledge from training data as guidance for prompt learning. However, existing methods still struggle with the generalization-adaptation (GA) trade-off dilemma: excessive reliance on generalized knowledge hinders adaptation to downstream tasks, while overemphasis on task-specific knowledge undermines the inherent generalization capabilities of pre-trained models. To address this issue, we propose a novel prompt learning method called Prompt Learning with Knowledge Regularization (PLKR). PLKR effectively mitigates the GA trade-off dilemma by offering greater flexibility in adapting to task-specific knowledge while minimizing the disruption of pre-trained knowledge. Specifically, we propose category-invariant and topology-invariant knowledge regularization to preserve generalized knowledge: the former enhances category-level discriminative capabilities while allowing flexible task-specific learning, and the latter maintains global topological stability during adaptation to new tasks. Through the proposed regularization, PLKR improves the performance on both base and new tasks. We evaluate the effectiveness of our approach on four representative tasks over 11 datasets. Experimental results show our method outperforms existing SOTA methods by a large margin.
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 11726-11735
Frequency-aware Correlation Discovering and Spatial Forgery Clue Distilling for Synthetic Image Detection
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 11726-11735
Open AccessFrequency-aware Correlation Discovering and Spatial Forgery Clue Distilling for Synthetic Image Detection
- 12 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36 (3) , 2757-2770
Spatial–Temporal Clue Reasoning Chain for Long Video Question Answering
Existing Video Question Answering (VideoQA) methods face tremendous challenges when dealing with longer videos. On the one hand, long videos contain rich and diverse information at different spatial-temporal scales, requiring the VideoQA model to mine multi-granularity visual clues conditioned on the specific question. On the other hand, the model needs to effectively reason the spatial-temporal relationships of complex events that span large temporal intervals. To address the above challenges, we propose a novel spatial-temporal clue reasoning chain to effectively reason multi-granularity visual clues and predict the answer in complex events. Specifically, we extracts question-related visual clues from video clips, frames and regions in a “step-by-step" manner. Further, we propose an asynchronous clue selector with self-excitation loss to perform more effective selections by (1) facilitating the identification of question-related visual clues and (2) enhancing the diversity of selections. Next, we introduce cross-modal fusion between the question semantics and the selected multi-granularity visual clues to learn the detailed interactions and then predict the answer. Extensive experiments on multiple VideoQA datasets demonstrate the effectiveness of our method.
- 12 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 36 (3) , 2757-2770
Spatial–Temporal Clue Reasoning Chain for Long Video Question Answering
Existing Video Question Answering (VideoQA) methods face tremendous challenges when dealing with longer videos. On the one hand, long videos contain rich and diverse information at different spatial-temporal scales, requiring the VideoQA model to mine multi-granularity visual clues conditioned on the specific question. On the other hand, the model needs to effectively reason the spatial-temporal relationships of complex events that span large temporal intervals. To address the above challenges, we propose a novel spatial-temporal clue reasoning chain to effectively reason multi-granularity visual clues and predict the answer in complex events. Specifically, we extracts question-related visual clues from video clips, frames and regions in a “step-by-step" manner. Further, we propose an asynchronous clue selector with self-excitation loss to perform more effective selections by (1) facilitating the identification of question-related visual clues and (2) enhancing the diversity of selections. Next, we introduce cross-modal fusion between the question semantics and the selected multi-granularity visual clues to learn the detailed interactions and then predict the answer. Extensive experiments on multiple VideoQA datasets demonstrate the effectiveness of our method.
- 1 March 2025
- Published by Elsevier in Pattern Recognition
- Vol. 164
A Bayesian dual-pathway network for unsupervised domain adaptation
- 1 March 2025
- Published by Elsevier in Pattern Recognition
- Vol. 164
A Bayesian dual-pathway network for unsupervised domain adaptation
- Open Access
- 28 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 8306-8315
Domain Shared and Specific Prompt Learning for Incremental Monocular Depth Estimation
Incremental monocular depth estimation aims to continuously learn from new domains while maintaining their performance on old domains. The catastrophic forgetting problem is the key challenge when the model adapts the dynamic scene variations. Previous methods usually address this forgetting problem by storing raw samples from the old domain, allowing the model to review the knowledge of the old domain. However, due to the concerns of data privacy and security, our objective is to tackle the incremental monocular depth estimation problem in more stringent scenarios without the need for replaying samples. In this paper, we attribute the cross-domain catastrophic forgetting to the domain distribution shifts and continuous variations of depth space. To this end, we propose Domain Shared and Specific Prompt Learning (DSSP) for incremental monocular depth estimation. In detail, to alleviate the domain distribution shift, complementary domain prompt is designed to learn the domain-shared and domain-specific knowledge which are optimized by the inter-domain alignment and intra-domain orthogonal loss. To mitigate the depth space variations, we first introduce a pre-trained model to generate the domain-shared depth space. Then, we design S^2-Adapter that quantizes depth space variations with scale&shift matrices and converts the domain-shared depth space to domain-specific depth space. Our method achieves state-of-the-art performance under various scenarios such as different depth ranges, virtual and real, different weather conditions, and the few-shot incremental learning setting on 12 datasets. We will release the source codes and pre-trained models.
- 28 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 8306-8315
Open AccessDomain Shared and Specific Prompt Learning for Incremental Monocular Depth Estimation
Incremental monocular depth estimation aims to continuously learn from new domains while maintaining their performance on old domains. The catastrophic forgetting problem is the key challenge when the model adapts the dynamic scene variations. Previous methods usually address this forgetting problem by storing raw samples from the old domain, allowing the model to review the knowledge of the old domain. However, due to the concerns of data privacy and security, our objective is to tackle the incremental monocular depth estimation problem in more stringent scenarios without the need for replaying samples. In this paper, we attribute the cross-domain catastrophic forgetting to the domain distribution shifts and continuous variations of depth space. To this end, we propose Domain Shared and Specific Prompt Learning (DSSP) for incremental monocular depth estimation. In detail, to alleviate the domain distribution shift, complementary domain prompt is designed to learn the domain-shared and domain-specific knowledge which are optimized by the inter-domain alignment and intra-domain orthogonal loss. To mitigate the depth space variations, we first introduce a pre-trained model to generate the domain-shared depth space. Then, we design S^2-Adapter that quantizes depth space variations with scale&shift matrices and converts the domain-shared depth space to domain-specific depth space. Our method achieves state-of-the-art performance under various scenarios such as different depth ranges, virtual and real, different weather conditions, and the few-shot incremental learning setting on 12 datasets. We will release the source codes and pre-trained models.