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
- Guanghui LiuSuggested from co-authorship
- King Ngi NganSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works219 from public data
- VisDrone-DET2020: The Vision Meets Drone Object Detection in Image Challenge Results259
Based on the selected 29 robust detection methods, the experimental results comprehensively are discussed, which shows the effectiveness of ensemble learning and data augmentation in drone captured object detection.
- Beyond pixels: A comprehensive survey from bottom-up to semantic image segmentation and cosegmentation240
A comprehensive review of the recent progress in image segmentation, covering 190 publications, gives an overview of broad segmentation topics including not only the classic unsupervised methods, but also the recent weakly-/semi- supervised methods and the fully-super supervised methods.
- 225
- Key-Word-Aware Network for Referring Expression Image Segmentation200
A key-word-aware network, which contains a query attention model and a key-words-aware visual context model, which outperforms state-of-the-art methods on two referring expression image segmentation databases.
- 165
- 161
- 145
- 114
- A2RMNet: Adaptively Aspect Ratio Multi-Scale Network for Object Detection in Remote Sensing Images76
A novel end-to-end Adaptively Aspect Ratio Multi-Scale Network (A 2 RMNet) is proposed to solve the problem of easy to miss or incorrectly locate objects due to the various sizes and aspect ratios of objects.
- Subjective and Objective De-Raining Quality Assessment Towards Authentic Rain Image75
A de-raining quality assessment (DQA) database is created and a bi-directional feature embedding network (B-FEN) which integrates the features of global perception and local difference together is proposed which significantly outperforms many existing universal blind image quality assessment models.
- 73
- 72
- 67
- 61
- Towards Continual Egocentric Activity Recognition: A Multi-Modal Egocentric Activity Dataset for Continual Learning59
A multi-modal egocentric activity dataset for continual activity learning named UESTC-MMEA-CL is proposed and the results of egocentric activity recognition of three modalities separately and jointly on a base network architecture are reported.
- 59
- 56
- 55
- 54
- Visual and Textual Prior Guided Mask Assemble for Few-Shot Segmentation and Beyond49
The proposed PGMA-Net can solve bbox-level and cross-domain FSS, co-segmentation, zero-shot segmentation (ZSS) tasks, leading an any-shot segmentation framework capable of accommodating diverse weak or pixel annotations.
- 49
- Global and local semantics-preserving based deep hashing for cross-modal retrieval49
A large margin is enforced between similar hash codes and dissimilar hash codes from an inter-modal view to learn discriminative hash codes, which can well preserve local semantic structure and the global semantic structure can be preserved into the hash codes.
- A SYMMETRIC FULLY CONVOLUTIONAL RESIDUAL NETWORK WITH DCRF FOR ACCURATE TOOTH SEGMENTATION48
A symmetric full convolutional network with residual block and Dense Conditional Random Field (DCRF), which can achieve accurate segmentation automatically for tooth images and can not only strengthen feature propagation, but also boost feature reuse.
- 47
- No reference image quality assessment metric via multi-domain structural information and piecewise regression47
This paper develops a new local image representation which extracts the structural image information from both the spatial-frequency and spatial domains and develops an efficient piecewise regression method to capture the local distribution of the feature space.
- 46
- 44
- 43
- Language-Aware Fine-Grained Object Representation for Referring Expression Comprehension41
This work proposes a language-aware deformable convolution model (LDC) that adaptively samples a set of key points based on the image and language to represent objects and designs a bidirectional interaction model (BIM) that leverages a modified co-attention mechanism to build cross-modalbidirectional interactions to further improve the language and object representations.
- Exploring space–frequency co-occurrences via local quantized patterns for texture representation41
This paper proposes to explore space-frequency co-occurrences via local quantized patterns for texture representation via global thresholding, and achieves the best classification results compared with state-of-the-art approaches investigated.
- 40
- 38
- A New Local Transformation Module for Few-Shot Segmentation38
A new transformation module based on local cues, where the relationship of the local features is used for transformation is proposed, which can be used to replace the transformation module in the existing few-shot segmentation frameworks.
- 38
- 38
- Discriminative deep metric learning for asymmetric discrete hashing37
A novel discriminative deep metric learning approach for asymmetric discrete hashing (ADMH) approach for supervised hashing learning and extensive experiments show that the proposed asymmetric deep hashing method outperforms the existing hashing methods.
- 36
- 36
- 35
- Cosegmentation of multiple image groups35
Experimental results demonstrate that the proposed multi-group cosegmentation framework is able to discover more accurate object priors and outperform state-of-the-art single-group image cose segmentation methods.
- 34
- 34
- Misaligned RGB-Infrared Object Detection via Adaptive Dual-Discrepancy Calibration30
An adaptive dual-discrepancy calibration network (ADCNet) for misaligned RGB-Infrared object detection, including spatial discrepancy and domain-discREPancy calibration, which outperforms the baseline by 3.3% and 2.5% in mAP50 on the FLIR and misaligned M3FD datasets, respectively.
- 29
- 29
- An efficient and compact 3D local descriptor based on the weighted height image29
The experimental results indicate that the proposed WHI descriptor outperforms the state-of-the-art (SOTA) algorithms in terms of accuracy and efficiency and the compactness of the WHI is about six times more than that of the SOTA algorithms.
- Non-Homogeneous Haze Removal via Artificial Scene Prior and Bidimensional Graph Reasoning28
This paper proposes a Non-Homogeneous Haze Removal Network (NHRN) via artificial scene prior and bidimensional graph reasoning and demonstrates that the method achieves superior performance over many state-of-the-art algorithms for both the single image dehazing and hazy image understanding tasks.
- 26
- 26
- Remember and Reuse25
A dynamic Remember and Reuse (R&R) network, which efficiently performs the cross-task BIQA based on a novel relevance-aware incremental learning strategy and significantly outperforms many state-of-the-art methods.
- A New Deep Segmentation Quality Assessment Network for Refining Bounding Box Based Segmentation24
This paper uses the proposed blind SQA network to refine the traditional bounding box-based segmentation results, with the idea of selecting parameters that are adaptive to each image, and a two-step refinement method is proposed.
- Semi-supervised manifold-embedded hashing with joint feature representation and classifier learning24
A semi-supervised manifold embedding is explored to simultaneously optimize feature representation and classifier learning to make the learned binary codes optimal for classification.
- 23
- ASFlow: Unsupervised Optical Flow Learning With Adaptive Pyramid Sampling23
This work presents an unsupervised optical flow estimation method by proposing an adaptive pyramid sampling in the deep pyramid network, and proposes a Content-Aware Pooling module, which promotes local feature gathering by avoiding cross region pooling, so that the learned features become more representative.
- 23
- 22
- 21
- Forgetting to Remember: A Scalable Incremental Learning Framework for Cross-Task Blind Image Quality Assessment21
A scalable incremental learning framework that could sequentially conduct BIQA across multiple evaluation tasks with limited memory capacity is proposed and a scalable memory unit is developed by gradually and selectively pruning unimportant neurons from previously settled parameter subsets to suppress the unrestrained expansion of memory capacity in sequential tasks learning.
- Manifold-ranking embedded order preserving hashing for image semantic retrieval21
A novel unsupervised hashing approach, namely Manifold-Ranking Embedded Order Preserving Hashing (MREOPH), which introduces a manifold ranking loss and an order preserving loss to solve the issue of global topological structure preserving.
- 21
- 21
- Cross-Modal Cognitive Consensus Guided Audio–Visual Segmentation20
A Cross-modal Cognitive Consensus guided Network (C3N) is proposed to align the audio-visual semantics from the global dimension and progressively inject them into the local regions via an attention mechanism.
- 20
- Bird breed classification and annotation using saliency based graphical model20
A saliency based graphical model (GMS) is proposed, which can precisely annotate the object on the pixel level and can achieve better results compared with existing bird breed classification methods based on graphical model.
- 19
- 19
- 19
- 19
- 19
- 17
- 16
- 16
- 16
- 15
- 15
- 14
- 14
- 14
- 14
- 14
- On Multiple Image Group Cosegmentation14
Experimental results show that the proposed multi-group cosegmentation framework is able to discover more accurate object priors and significantly outperform state-of-the-art single-group image cose segmentation methods.
- 13
- 12
- RefCrowd: Grounding the Target in Crowd with Referring Expressions12
A Fine-grained Multi-modal Attribute Contrastive Network (FMAC) to deal with REF in crowd understanding, which outperforms existing state-of-the-art (SoTA) methods on the authors' RefCrowd dataset and existing REF datasets, and implements an end-to-end REF toolbox for the deeper research in multi- modal domain.
- 11
- Continual Egocentric Activity Recognition With Foreseeable-Generalized Visual–IMU Representations11
This work proposes a motivational optimization scheme to address the limited generalization caused by the modal imbalance, enabling foreseeable generalization in a visual–IMU multimodal network and introduces a robust representation estimation technique and pseudo representation (PR) generation strategy for continual learning.
- 11
- 11
- 11
- 11
- 10
- 10
- Boosting Scene Parsing Performance via Reliable Scale Prediction10
A novel fully-supervised Scale Prediction Model is proposed, which learns parsing scales by the strong scale supervision, which is automatically generated from the scene parsing ground truth without any extra manually annotation, and a Scale Parsing Net (SPNet) is designed, which segments each object on the scale predicted by the Scale Prediction model.
- 10
- 9
- 9
- A New Few-shot Segmentation Network Based on Class Representation9
A new class representation based few-shot segmentation framework, which firstly generates class activation map of unseen class based on the knowledge of existing classes, and then uses the map as foreground probability map to extract the foregrounds from query image.
- 9
- Learning With Noisy Low-Cost MOS for Image Quality Assessment via Dual-Bias Calibration8
This paper considers the LC-MOS as the noisy observation of LA-MOS and enforce the IQA model learned from LC-MOS to approach the unbiased estimation of LA-MOS, which is the first exploration of robust IQA model learning from noisy low-cost labels.
- 8
- 8
- A new single-view 3D pantograph reconstruction aided by prior CAD model8
A novel 3D pantograph reconstruction method based on the single-view scanned data and prior Computer-Aided Design (CAD) model that can reconstruct the complete 3D model with high accuracy has been successfully applied to the anomaly detection and visualization in the pantograph inspection.
- 8
- 8
- 8
- 7
- Cognition Transferring and Decoupling for Text-Supervised Egocentric Semantic Segmentation7
A Cognition Transferring and Decoupling Network (CTDN) is proposed that first learns the egocentric wearer-object relations via correlating the image and text and a Cognition Transferring Module (CTM) is developed to distill the cognitive knowledge from the large-scale pre-trained model to the authors' model for recognizing egocentric objects with various semantics.
- 7
- 7
- Category boundary re-decision by component labels to improve generation of class activation map7
This work proposes to use a set of lower-level labels, called component labels, to refine the classification boundary and provide more supervision to force the classification network to learn more local features of the target.
- 7
- 7
- 7
- 7
- 6
- Class Activation Map Generation by Multiple Level Class Grouping and Orthogonal Constraint6
This paper tries to capture the semantic relationships between classes to form different semantic levels of classification models, and hierarchical clustering based on class relationships is used to form hierarchical clusters results, where the clustering levels are treated as semantic levels to form the classification models.
- 6
- 5
- 5
- 5
- 5
- Learning Offset Probability Distribution for Accurate Object Detection5
A distance-aware offset bin classification network optimized with multiple binary cross entropy losses to learn various offset probability distribution, including single label distribution and distance-aware label distribution is proposed.
- 5
- Multi-Scale Shape Adaptive Network for Raindrop Detection and Removal from a Single Image5
This paper establishes a large-scale dataset named RaindropCityscapes, which includes 11,583 pairs of raindrop and raindrop-free images, covering a wide variety of raindrops and background scenarios, and proposes a two-branch Multi-scale Shape Adaptive Network (MSANet) to detect and remove diverse raindrops.
- 5
- 5
- Unsupervised Ego- and Exo-centric Dense Procedural Activity Captioning via Gaze Consensus Adaptation4
A Gaze Consensus-guided Ego-Exo Adaptation Network (GCEAN) that injects the gaze information into the learned representations for the fine-grained Ego-Exo alignment and adopts hierarchical gaze-guided consistency losses to construct gaze consensus for the explicit temporal and spatial adaptation between the source and target views.
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- Hierarchical class grouping with orthogonal constraint for class activation map generation3
Experimental results on the PASCAL VOC 2012 dataset show the superior performance of the proposed CAM generation method, which attributes such incompleteness extraction to the finite discriminative cues within a single classification model.
- 3
- Salient Object Detection and Segmentation via Ultra-Contrast3
This paper proposes a salient object detection principle from existing contrast based methods: dissimilarity produces contrast, while contrast leads to saliency, and proposes a generalized framework to detect complex salient objects.
- 3
- On the Adversarial Robustness of Learning-Based Image Compression Against Rate-Distortion Attacks2
This paper considers the arbitrary nature of the attack direction and the uncontrollable compression ratio faced by adversaries, and presents two practical rate-distortion attack paradigms, i.e., Specific-ratio Rate-Distortion Attack (SRDA) and Agnostic-ratio Rate-Distortion Attack (ARDA), and is the first to conduct joint rate-distortion attacks on LIC algorithms.
- Distribution-Level Memory Recall for Continual Learning: Preserving Knowledge and Avoiding Confusion2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 1
- 1
- 1
- 1
- Your Demands Deserve More Bits: Referring Semantic Image Compression at Ultra-low Bitrate1
This work proposes a novel Referring Semantic Image Compression framework, which achieves flexible customized compression according to user demands, which better balance the local fidelity, global realism, semantic alignment, and bit overhead.
- 1
- 1
- No Re-Train, More Gain: Upgrading Backbones with Diffusion model for Pixel-Wise and Weakly-Supervised Few-Shot Segmentation1
DiffUp, a novel FSS method that conceptualizes the FSS task as a conditional generative problem using a diffusion process, and incorporates an uncertainty-aware information fusion module that harmonizing the variability across zero-shot, one-shot and many-shot scenarios.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- DFR: A Decompose-Fuse-Reconstruct Framework for Multi-Modal Few-Shot Segmentation–
A novel framework that addresses the fundamental challenge of effectively utilizing multi-modal guidance in few-shot segmentation (FSS), and leverages the Segment Anything Model to systematically integrate visual, textual, and audio modalities for enhanced semantic understanding.
- –
- –
- –
- –
- –
- –
- –
- –
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 2026-10-16. 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.