Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Yanning ZhangSuggested from co-authorship
- Nanning ZhengSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works42 from public data
- Dual Modality Prompt Tuning for Vision-Language Pre-Trained Model118
This article proposes a novel dual-modality prompt tuning (DPT) paradigm through learning text and visual prompts simultaneously and proposes a class-aware visual prompt tuning (CAVPT) scheme, which is further proposed in this DPT.
- Towards Robust Event-guided Low-Light Image Enhancement: A Large-Scale Real-World Event-Image Dataset and Novel Approach81
This work designs the multiscale holistic fusion branch to extract holistic structural and textural information from both events and images to ensure robustness against variations in the regional illumination and noise, and introduces a Signal-to-Noise-Ratio (SNR)-guided regional feature selection.
- Frequency-Guided Spatial Adaptation for Camouflaged Object Detection64
A novel frequency-guided spatial adaptation method for COD task that outperforms 26 state-of-the-art methods with large margins by transforming the input features of the adapter into frequency domain.
- Attend to the Difference: Cross-Modality Person Re-Identification via Contrastive Correlation63
A dual-path cross-modality feature learning framework which preserves intrinsic spatial structures and attends to the difference of input cross- modality image pairs, which outperforms state-of-the-art algorithms by a large margin with both full and simplified evaluation modes.
- Pseudo Labeling Methods for Semi-Supervised Semantic Segmentation: A Review and Future Perspectives57
This paper provides a comprehensive review and analysis of various pseudo-labeling methods and their applications in semi-supervised semantic segmentation (SSSS) from multiple angles and evaluates the performance of cutting-edge methods on public datasets.
- MS-DETR: Multispectral Pedestrian Detection Transformer With Loosely Coupled Fusion and Modality-Balanced Optimization48
This paper proposes an instance-aware modality-balanced optimization strategy, which preserves visible and thermal decoder branches and aligns their predicted slots through an instance-wise dynamic loss, and shows superior performance on the challenging KAIST, CVC-14 and LLVIP benchmark datasets.
- An adversarial human pose estimation network injected with graph structure48
A novel generative adversarial network (GAN) is designed to improve the localization accuracy of visible joints when some joints are invisible to produce accurate human pose prediction.
- 42
- Kiwi-Edit: Versatile Video Editing via Instruction and Reference Guidance36
A scalable data generation pipeline is introduced that transforms existing video editing pairs into high-fidelity training quadruplets, leveraging image generative models to create synthesized reference scaffolds, and a unified editing architecture is proposed, Kiwi-Edit, that synergizes learnable queries and latent visual features for reference semantic guidance.
- 36
- 31
- 26
- CoLeCLIP: Open-Domain Continual Learning via Joint Task Prompt and Vocabulary Learning23
A novel approach is introduced, termed CoLeCLIP, which learns an open-domain CL model based on CLIP, which achieves new state-of-the-art performance for open-domain CL under both task- and class-incremental learning (CIL) settings.
- Ground-to-Aerial Person Search: Benchmark Dataset and Approach22
This work constructs a large-scale dataset for Ground-to-Aerial Person Search, named G2APS, which contains 31,770 images of 260,559 annotated bounding boxes for 2,644 identities appearing in both of the UAVs and ground surveillance cameras, and proposes a simple yet effective knowledge distillation scheme on the head of the ReID network.
- Non-Exemplar Class-Incremental Learning by Random Auxiliary Classes Augmentation and Mixed Features19
An effective non-exemplar method called RAMF consisting of Random Auxiliary classes augmentation and Mixed Features, where the method outperforms the state-of-the-art non-exemplar methods and is comparable to high-performance replay-based methods.
- New Insights on Relieving Task-Recency Bias for Online Class Incremental Learning18
An Adaptive Focus Shifting algorithm (AFS), which dynamically adjusts focus to ambiguous samples and non-target logits in model learning, and introduces a virtual knowledge distillation to promote plasticity.
- Text-based Person Search in Full Images via Semantic-Driven Proposal Generation17
This work proposes a new end-to-end learning framework which jointly optimize the pedestrian detection, identification and visual-semantic feature embedding tasks and achieves the state-of-the-art performance.
- Pose-and-illumination-invariant face representation via a triplet-loss trained deep reconstruction model16
A new residual based deep face reconstruction neural network to extract discriminative pose-and-illumination-invariant features and achieves 83.8% average recognition accuracy on MultiPIE face dataset which is competitive to the state-of-the-art face recognition methods.
- 15
- 10
- Cross-View Person Identification by Matching Human Poses Estimated With Confidence on Each Body Joint10
A new metric of confidence to the 3D human pose estimation is introduced and it is shown that the combination of the inaccurately estimated human pose and the inferred confidence metric can be used to boost the CVPI performance---the estimated pose information can be integrated to the appearance and motion features to achieve the new state-of-the-art C VPI performance.
- 6
- Local-enhanced Interaction for Temporal Moment Localization5
A multi-branches interaction model for temporal moment localization via language that encoded the query sentence and video are encoded into multiple feature embeddings over several semantic sub-spaces, which demonstrates the effectiveness of the proposed model.
- 4
- An Adaptive Viewpoint Transformation Network for 3D Human Pose Estimation4
Experiments on Human3.6M and MPII datasets show that the proposed adaptive viewpoint transformation can improve the performance of 3D human pose estimation.
- 4
- 3
- Event-Guided Consistent Video Enhancement with Modality-Adaptive Diffusion Pipeline2
This paper designs a modality-agnostic conditioning mechanism based on a diffusion pipeline that dynamically adjusts the contribution of each modality according to sensor-specific characteristics, and introduces a modality-adaptive guidance rescaling that dynamically adjusts the contribution of each modality according to sensor-specific characteristics.
- Sparkle: Realizing Lively Instruction-Guided Video Background Replacement via Decoupled Guidance1
This paper designs a scalable pipeline that generates foreground and background guidance in a decoupled manner with strict quality filtering, and introduces Sparkle, a dataset of ~140K video pairs spanning five common background-change themes, alongside Sparkle-Bench, the largest evaluation benchmark tailored for background replacement to date.
- 1
- DMAT: A Dynamic Mask-Aware Transformer for Human De-occlusion1
A dynamic mask-aware transformer (DMAT), which dynamically augments information from human regions and weakens that from occlusion, and designs an expanded convolution head with enlarged kernels to enhance token representation.
- 1
- Intention-Based Human Robot Collaboration1
In this framework, the behavior of robots is based on the actively recognition and understanding of the dynamic scene and human intention, which is one of the core character of the coexisting robot.
- –
- –
- –
- –
- –
- –
- –
- Boosting Multi-Modal Alignment: Geometric Feature Separation for Class Incremental Learning–
Inspired by neural collapse theory, this work introduces a multi-modal alignment strategy that optimizes the geometric structure of both visual and textual features across different classes in CIL, and achieves state-of-the-art performance.
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.