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
- Hassan ForooshSuggested from co-authorship
- Mubarak A. ShahSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works587 from public data
- A review of uncertainty quantification in deep learning: Techniques, applications and challenges2,893
This study reviews recent advances in UQ methods used in deep learning and investigates the application of these methods in reinforcement learning (RL), and outlines a few important applications of UZ methods.
- Single Image Dehazing via Multi-scale Convolutional Neural Networks1,759
A multi-scale deep neural network for single-image dehazing by learning the mapping between hazy images and their corresponding transmission maps by combining a coarse-scale net which predicts a holistic transmission map based on the entire image, and a fine-scale network which refines results locally.
- Gated Fusion Network for Single Image Dehazing973
An efficient algorithm to directly restore a clear image from a hazy input using an end-to-end trainable neural network that consists of an encoder and a decoder is proposed.
- Single Image Super-Resolution via a Holistic Attention Network857
A new holistic attention network (HAN) is proposed, which consists of a layer attention module (LAM) and a channel-spatial attention Module (CSAM) to model the holistic interdependencies among layers, channels, and positions.
- 767
- 739
- 566
- 562
- 508
- Chat-UniVi: Unified Visual Representation Empowers Large Language Models with Image and Video Understanding506
This work introduces Chat-UniVi, a Unified Vision-language model capable of comprehending and engaging in conver-sations involving images and videos through a unified visual representation that outperforms even existing methods exclusively de-signed for either images or videos.
- Consistent and Specific Multi-View Subspace Clustering450
A novel multi-view subspace clustering method (CSMSC), where consistency and specificity are jointly exploited for subspace representation learning, and forms the multi-View self-representation property using a shared consistent representation and a set of specific representations, which better fits the real-world datasets.
- Cluster-Based Co-Saliency Detection427
This paper introduces a new cluster-based algorithm for co-saliency detection that is mostly bottom-up without heavy learning, and outperforms most the state-of-the-art saliency detection methods.
- Deep People Counting in Extremely Dense Crowds388
This paper proposes an end-to-end deep convolutional neural networks (CNN) regression model for counting people of images in extremely dense crowds that achieves superior performance than the state-of-the-arts in term of the mean and variance of absolute difference.
- Single Image Dehazing via Multi-scale Convolutional Neural Networks with Holistic Edges382
A multi-scale deep neural network for single image dehazing by learning the mapping between hazy images and their transmission maps by combining a coarse-scale net which predicts a holistic transmission map based on the entire image, and a fine-scale network which refines dehazed results locally.
- 380
- Single Image Deraining: A Comprehensive Benchmark Analysis371
A comprehensive study and evaluation of existing single image deraining algorithms, using a new large-scale benchmark consisting of both synthetic and real-world rainy images, and provides a rich variety of criteria for dehazing algorithm evaluation.
- Depth Enhanced Saliency Detection Method365
This paper proposes a saliency detection method using the additional depth information and saliency cues are provided to follow the laws of the visually salient stimuli in both color and depth spaces.
- 364
- Dense Attention Fluid Network for Salient Object Detection in Optical Remote Sensing Images350
An end-to-end Dense Attention Fluid Network (DAFNet) for SOD in optical RSIs that significantly outperforms the existing state-of-the-art SOD competitors is proposed and a new and challenging optical RSI dataset is constructed that contains 2,000 images with pixel-wise saliency annotations.
- ComDefend: An Efficient Image Compression Model to Defend Adversarial Examples340
An end-to-end image compression model to defend adversarial examples: ComDefend, which outperforms the state-of-the-art defense methods, and is consistently effective to protect classifiers against adversarial attacks.
- 296
- 281
- 281
- Transferable Adversarial Attacks for Image and Video Object Detection280
The proposed method is based on the Generative Adversarial Network (GAN) framework, where it combines the high-level class loss and low-level feature loss to jointly train the adversarial example generator, and can efficiently generate image and video adversarial examples that have better transferability.
- Logit Standardization in Knowledge Distillation271
The proposed Z-score pre-process enables student to focus on essential logit relations from teacher rather than requiring a magnitude match, and can improve the performance of existing logit-based distillation methods.
- 236
- Saliency Detection for Stereoscopic Images Based on Depth Confidence Analysis and Multiple Cues Fusion235
A measure to evaluate the reliability of depth map, and use it to reduce the influence of poor depth map on saliency detection, and two saliency maps are integrated into a final saliency map through weighted-sum method according to their importance.
- 231
- CIR-Net: Cross-Modality Interaction and Refinement for RGB-D Salient Object Detection229
A convolutional neural network (CNN) model, named CIR-Net, based on the novel cross-modality interaction and refinement is presented, which outperforms the state-of-the-art saliency detectors both qualitatively and quantitatively.
- 213
- Semantic Community Identification in Large Attribute Networks211
A novel nonnegative matrix factorization (NMF) model with two sets of parameters, the community membership matrix and community attribute matrix is proposed and the use of node attributes improves upon community detection and provides a semantic interpretation to the resultant network communities.
- 207
- Dual Quaternion Knowledge Graph Embeddings204
The core of DualE lies a specific design of dual-quaternion-based multiplication, which universally models relations as the compositions of a series of translation and rotation operations.
- 202
- 197
- LAS-AT: Adversarial Training with Learnable Attack Strategy196
A novel framework for adversarial training is proposed by introducing the concept of “learnable attack strategy”, dubbed LAS-AT, which learns to automatically produce attack strategies to improve the model robustness.
- 185
- 172
- Matching-CNN meets KNN: Quasi-parametric human parsing169
This work aims to develop a new solution with the advantages of both methodologies, namely supervision from annotated data and the flexibility to use newly annotated (possibly uncommon) images, and present a quasi-parametric human parsing model.
- 168
- 167
- 159
- 149
- Tensorized Multi-view Subspace Representation Learning148
A novel algorithm termed as Tensorized Multi-view Subspace Representation Learning is established, which models elegantly the complementary information among different views, reduces redundancy of subspace representations, and then improves the accuracy of subsequent tasks.
- BadCLIP: Dual-Embedding Guided Backdoor Attack on Multimodal Contrastive Learning145
The BadCLIP attack is introduced, which is resistant to backdoor detection and model fine-tuning defenses and draws motivations from the perspective of the Bayesian rule and proposes a dual-embedding guided framework for backdoor attacks.
- 137
- Improved Techniques for Optimization-Based Jailbreaking on Large Language Models134
Improved techniques for optimization-based jailbreaks like GCG are presented and demonstrated that they can help GCG outperform state-of-the-art jailbreaking attacks and achieve nearly 100% attack success rate.
- 134
- 132
- Fusing two-stream convolutional neural networks for RGB-T object tracking128
A novel Convolutional Neural Network (ConvNet) architecture, including a two-stream ConvNet and a FusionNet, is proposed to achieve adaptive fusion of different source data for robust RGB-T tracking.
- 127
- An Iterative Co-Saliency Framework for RGBD Images123
An iterative RGBD co-saliency framework is proposed, which utilizes the existing single saliency maps as the initialization, and generates the final RGBD Co-Saliency map by using a refinement-cycle model.
- 122
- Adv-watermark: A Novel Watermark Perturbation for Adversarial Examples121
A novel watermark perturbation for adversarial examples (Adv-watermark) which combines image watermarking techniques and adversarial example algorithms and outperforms the state-of-the-art attack methods.
- 116
- Topology Optimization based Graph Convolutional Network114
A novel Topology Optimization based Graph Convolutional Networks (TO-GCN) to fully utilize the potential information by jointly refining the network topology and learning the parameters of the FCN is proposed.
- 112
- Learning With Multiclass AUC: Theory and Algorithms109
This paper proposes an empirical surrogate risk minimization framework to approximately optimize the M metric, and proposes acceleration methods for three popular surrogate loss functions, including the exponential loss, squared loss, and hinge loss, to speed up loss and gradient evaluations.
- Parallel Rectangle Flip Attack: A Query-based Black-box Attack against Object Detection108
Extensive experiments demonstrate that the proposed Parallel Rectangle Flip Attack (PRFA) can effectively and efficiently attack various popular object detectors, including anchor-based and anchor- free, and generate transferable adversarial examples.
- Efficient Adversarial Attacks for Visual Object Tracking105
This paper presents an end-to-end network FAN (Fast Attack Network) that uses a novel drift loss combined with the embedded feature loss to attack the Siamese network based trackers.
- HSCS: Hierarchical Sparsity Based Co-saliency Detection for RGBD Images104
A novel co-saliency detection method for RGBD images based on hierarchical sparsity reconstruction and energy function refinement is proposed, which outperforms the state-of-the-art algorithms both qualitatively and quantitatively.
- 99
- Video Deblurring via Semantic Segmentation and Pixel-Wise Non-linear Kernel99
This paper analyzes the relationship between motion blur trajectory and optical flow, and presents a novel pixel-wise non-linear kernel model to account for motion blur, which describes complex motion blur more effectively.
- Defending against Model Stealing via Verifying Embedded External Features97
Experimental results demonstrate that the method is effective in detecting different types of model stealing simultaneously, even if the stolen model is obtained via a multi-stage stealing process.
- Deep Non-Blind Deconvolution via Generalized Low-Rank Approximation96
The authors' analysis shows that the estimated decomposed matrices contain the most essential information of the input kernel, which ensures the proposed network to handle various blurs in a unified framework and generate high-quality deblurring results.
- Survey of recent progress in semantic image segmentation with CNNs96
This work surveys the powerful CNNs and novel elaborate layers, structures and strategies, especially including those that have achieved the state-of-the-art results on the Pascal VOC 2012 semantic segmentation challenge and proposes several possible directions and approaches to incorporate existing effective methods as components to enhance CNNs for the segmentation of specific semantic objects.
- Face Super-Resolution Guided by 3D Facial Priors89
This paper proposes a novel face super-resolution method that explicitly incorporates 3D facial priors which grasp the sharp facial structures and is the first to explore 3D morphable knowledge based on the fusion of parametric descriptions of face attributes.
- 89
- Video dehazing with spatial and temporal coherence89
A new framework for video dehazing, the process of restoring the visibility of the videos taken under foggy scenes, which builds upon techniques in single image dehaze, optical flow estimation and Markov random field to improve the temporal and spatial coherence of the dehazed video.
- 87
- 86
- A Survey of Orthogonal Moments for Image Representation: Theory, Implementation, and Evaluation82
A comprehensive survey of the orthogonal moments for image representation, covering recent advances in fast/accurate calculation, robustness/invariance optimization, definition extension, and application.
- 81
- 81
- Boosting Fast Adversarial Training With Learnable Adversarial Initialization79
Experimental evaluations on four benchmark databases demonstrate the superiority of the proposed method over state-of-the-art fast AT methods, as well as comparable robustness to advanced multi-step AT methods.
- 79
- Degradation-Guided One-Step Image Super-Resolution with Diffusion Priors78
This work introduced a novel one-step SR model, which significantly addresses the efficiency issue of diffusion-based SR methods and tailor a novel training pipeline by introducing an online negative sample generation strategy.
- 77
- A Survey on Transferability of Adversarial Examples across Deep Neural Networks76
This survey explores the landscape of the adversarial transferability of adversarial examples and categorizes existing methodologies to enhance adversarial transferability, and discusses the fundamental principles guiding each approach.
- Identifying overlapping communities as well as hubs and outliers via nonnegative matrix factorization76
A novel model where vertices are measured by their centrality in communities and defined as an optimization problem, calculated by nonnegative matrix factorization is proposed, which demonstrates its ability of identifying overlapping communities, hubs and outliers and validate its superior performance in terms of clustering quality.
- Fake Colorized Image Detection75
Two simple yet effective detection methods for fake colorized images are proposed: Histogram-basedfake colorized image detection and feature encoding-based fake colorize image detection, which exhibit a decent performance against multiple state-of-the-art colorization approaches.
- 73
- A Comprehensive Benchmark Analysis of Single Image Deraining: Current Challenges and Future Perspectives71
A comprehensive study and evaluation of existing single image deraining algorithms, using a new large-scale benchmark consisting of both synthetic and real-world rainy images of various rain types, which indicates the gap between the achievable performance on synthetic rainy images and the practical demand on real-world images.
- 70
- 70
- Improving Fast Adversarial Training With Prior-Guided Knowledge69
A positive prior-guided adversarial initialization to prevent overfitting by improving adversarial example quality without extra training time is proposed and can effectively improve the model's adversarial robustness in white-box attack scenarios.
- Visual Content Privacy Protection: A Survey69
A framework that encompasses various concerns and solutions for visual privacy is proposed, which allows for a macro understanding of privacy concerns from a comprehensive level and divides privacy protection into three categories, based on computer vision adversary, based on human vision (HV) adversary, and based on CV & HV adversary.
- 69
- 67
- 67
- 66
- 66
- 65
- 64
- 64
- Generating Transferable 3D Adversarial Point Cloud via Random Perturbation Factorization62
This paper proposes a simple yet effective attack method to generate more transferable adversarial 3D point clouds by combining the loss of perturbation with its random factorization and conducts experiments to verify its effectiveness in increasing transferability while preserving high efficiency.
- 62
- Revisiting Backdoor Attacks against Large Vision-Language Models from Domain Shift61
A multimodal attribution backdoor attack (MABA) is proposed that injects domain-agnostic triggers into critical areas using attributional interpretation and boosts the attack success rate of generalization over the unimodal attack.
- Research on axial bearing capacity of rectangular concrete-filled steel tubular columns based on artificial neural networks61
Parametric study was conducted using ANN1 and ANN2 which indicates that effect law of basic parameters of columns on the axial bearing capacity of rectangular CFT columns differs from design codes, and provides convincing design reference to rectangular C FT columns.
- Graph Attention Topic Modeling Network60
A new method to overcome the overfitting issue of pLSI is provided by using the amortized inference with word embedding as input, instead of the Dirichlet prior in LDA.
- SHIELD: an evaluation benchmark for face spoofing and forgery detection with multimodal large language models59
This study introduces a benchmark, SHIELD, to evaluate MLLMs for face spoofing and forgery detection, and proposes a novel multi-attribute chain of thought (MA-COT) paradigm for describing and judging various task-specific and task-irrelevant attributes of face images.
- 59
- 59
- 59
- Robust Foreground Detection Using Smoothness and Arbitrariness Constraints59
This paper forms the problem in a unified framework from a probabilistic perspective, and design an effective algorithm to seek the optimal solution, and Experimental results on both synthetic and real data demonstrate the clear advantages of the method compared to the state of the art alternatives.
- 59
- Hierarchical Graph Interaction Transformer With Dynamic Token Clustering for Camouflaged Object Detection58
A hierarchical graph interaction network termed HGINet is proposed, which is capable of discovering imperceptible objects via effective graph interaction among the hierarchical tokenized features via effective graph interaction among the hierarchical tokenized features.
- Graph Neural Networks Beyond Compromise Between Attribute and Topology57
A novel GNN, i.e., Graph Neural Network Beyond Compromise (GNN-BC), is proposed to iteratively updates the representations of topology and attribute by simultaneously capturing semantic information and removing the common information, and the final representation is the concatenation of them.
- DM2C: Deep Mixed-Modal Clustering57
This paper presents an adversarial learning framework for clustering with mixed-modal data that learns the mappings across individual modal spaces by virtue of cycle-consistency and could easily unify all the samples into a single modal space and perform the clustering.
- Camera calibration and light source orientation from solar shadows56
This paper describes a method for recovering camera parameters from perspective views of daylight shadows in a scene, given only minimal geometric information determined from the images, which uses cast shadows by the sun and requires no measurements of any distance or angle in the 3D world.
- Hide in Thicket: Generating Imperceptible and Rational Adversarial Perturbations on 3D Point Clouds54
A novel shape-based adversarial attack method, HiT-ADV, which initially conducts a two-stage search for attack regions based on saliency and imperceptibility scores, and then adds deformation perturbations in each attack region using Gaussian kernel functions, which is extendable to physical attack.
- 54
- 54
- 54
- Camera calibration and geo-location estimation from two shadow trajectories54
It is shown that camera parameters, latitude, longitude and shadow casting objects' relative height ratios can be estimated from two observed shadow trajectories and the design of an analemmatic sundial is used to get the shadow conic and furthermore recover the camera's geographical location.
- Identity-Preserving Face Anonymization via Adaptively Facial Attributes Obfuscation53
A face anonymization framework that could obfuscate visual appearance while preserving the identity discriminability and can achieve a balance between recognition utility and appearance anonymizing by modifying different numbers of facial attributes according to pratical demands.
- Pre-trained Trojan Attacks for Visual Recognition52
The Pre-trained Trojan attack is proposed, which embeds backdoors into a PVM, enabling attacks across various downstream vision tasks, and addresses the issue of shortcut connections by introducing a context-free learning pipeline for poison training.
- Implicit Feedbacks are Not Always Favorable: Iterative Relabeled One-Class Collaborative Filtering against Noisy Interactions51
A novel iteratively relabeling framework to jointly mitigate the noise in both observed and unobserved interactions is proposed, which exploits the self-training principle to dynamically generate pseudo labels for user preferences.
- AIParsing: Anchor-Free Instance-Level Human Parsing50
An instance-level human parsing network which is anchor-free and solvable on a pixel level, which achieves the best global-level and instance- level performance over state-of-the-art one-stage top-down alternatives.
- Modeling Multi-Task Model Merging as Adaptive Projective Gradient Descent49
This work frames model merging as a constrained optimization problem, minimizing the gap between the merged model and individual models, subject to the constraint of retaining shared knowledge, and solves it via adaptive projective gradient descent via adaptive projective gradient descent.
- 49
- 49
- Good match exploration using triangle constraint48
This paper presents a novel method for addressing the problem of finding more good feature pairs between images, which is one of the most fundamental tasks in computer vision and pattern recognition and significantly improves both the number of correct matches and the matching score.
- Poisoned Forgery Face: Towards Backdoor Attacks on Face Forgery Detection47
A novel and previously unrecognized threat in face forgery detection scenarios caused by backdoor attack by embedding backdoors into models and incorporating specific trigger patterns into the input, attackers can deceive detectors into producing erroneous predictions for forged faces.
- SA-Attack: Improving Adversarial Transferability of Vision-Language Pre-training Models via Self-Augmentation47
This work proposes a self-augment-based transfer attack method, termed SA-Attack, which applies different data augmentation methods to the image modality and text modality, respectively, with the aim of improving the adversarial transferability of the generated adversarial images and texts.
- 47
- 47
- 47
- 46
- 46
- 46
- 45
- 45
- Geometric Occlusion Analysis in Depth Estimation Using Integral Guided Filter for Light-Field Image45
This paper analyzes the geometric relationship between ASIs and reference sub-aperture images, and proves the occlusion boundary similarity, and shows that some points in ASI have higher reliability than other points for depth calculation.
- 45
- SurgVLM: A Large Vision-Language Model and Systematic Evaluation Benchmark for Surgical Intelligence44
This work proposes SurgVLM, one of the first large vision-language foundation models for surgical intelligence, where this single universal model can tackle versatile surgical tasks and builds upon Qwen2.5-VL, which is built upon Qwen2.5-VL and undergoes instruction tuning to 10+ surgical tasks.
- OT-Attack: Enhancing Adversarial Transferability of Vision-Language Models via Optimal Transport Optimization44
The proposed method formulates the features of image and text sets as two distinct distributions and employs optimal transport theory to determine the most efficient mapping between them, which informs the generation of adversarial examples to effectively counteract the overfitting issues.
- 44
- Heterogeneous Graph Information Bottleneck43
A novel Heterogeneous Graph Information Bottleneck (HGIB) is proposed to implement the consensus hypothesis in an unsupervised manner and significantly outperforms most semi-supervised SOTA methods based on complementary assumption.
- Pyramid Architecture Search for Real-Time Image Deblurring42
A novel deblurring method, dubbed PyNAS (pyramid neural architecture search network), towards automatically designing hyper-parameters including the scales, patches, and standard cell operators, and uses a path-level binarization mechanism for multi-scale cell searching to save the memory consumption.
- 42
- 42
- 41
- MOVE: Effective and Harmless Ownership Verification via Embedded External Features40
An effective and harmless model ownership verification (MOVE) to defend against different types of model stealing simultaneously, without introducing new security risks, is proposed.
- 40
- MaxMatch: Semi-Supervised Learning With Worst-Case Consistency40
This article presents a generalization bound for SSL, and derives an SSL objective that minimizes the largest inconsistency between an original unlabeled sample and its multiple augmented variants, and provides a simple but effective algorithm to solve the proposed minimax problem.
- 40
- A Hybrid Transformer-Mamba Network for Single Image Deraining39
A novel dual-branch hybrid Transformer-Mamba network, denoted as TransMamba, aimed at effectively capturing long-range rain-related dependencies and developing a spectral coherence loss to better reconstruct innate signal-level relations within clean images.
- 39
- Why Do Attributes Propagate in Graph Convolutional Neural Networks?39
This paper aims at providing a novel explanation to the question of "Why do attributes propagate in GCNNs?''. which not only gives the essence of the oversmoothing, but also illustrates why the GCN extensions, including multi-scale GCN and GCN with initial residual, can improve the performance.
- Facial Kinship Verification: A Comprehensive Review and Outlook38
A comprehensive review of the problem of Facial Kinship Verification, including problem definition, challenges, applications, benchmark datasets, a taxonomy of existing methods, and state-of-the-art performance is conducted.
- Co-Saliency Detection via Base Reconstruction38
This paper addresses the co-saliency detection to a reconstruction problem: the foreground could be well reconstructed by using the reconstruction bases, which are extracted from each image and have the similar appearances in the feature space.
- MIFT: A Mirror Reflection Invariant Feature Descriptor38
The structure of MIFT is analyzed and it is shown how MIFT outperforms SIFT in the context of mirror reflection while performs as well as SIFT when there is no mirror reflection.
- 37
- Unlearning Backdoor Threats: Enhancing Backdoor Defense in Multimodal Contrastive Learning via Local Token Unlearning36
This paper strengthens the backdoor shortcuts to discover suspicious samples through overfitting training prioritized by weak similarity samples and introduces an innovative token-based localized forgetting training regime that ensures a minimal success rate for attacks, and preserves the model's high clean accuracy.
- 36
- 36
- Self-calibration from turn-table sequences in presence of zoom and focus36
A linear method that works under restrictive conditions on camera internal parameters, the solution of which can be used as the starting point of the iterative non-linear method with looser constraints, and the 3D structure of the object is recovered and the camera parameters are further refined simultaneously.
- Multimodal Unlearnable Examples: Protecting Data against Multimodal Contrastive Learning35
This paper proposes Multi-step Error Minimization (MEM), a novel optimization process for generating unlearnable examples that extends the Error-Minimization framework to optimize both image noise and an additional text trigger, thereby enlarging the optimized space and effectively misleading the model to learn the shortcut between the noise features and the text trigger.
- Imitated Detectors35
It is revealed for the first time that black-box victim object detectors can be easily replicated without knowing the model structure and training data and a teacher-student framework named Imitated Detector is proposed to transfer the knowledge of the victim model to the imitated model.
- When to Learn What35
A robust deep subspace clustering framework based on the principle of human cognitive process is proposed, and the easinesses of samples are measured dynamically so that the method could gradually utilize instances from easy to more complex ones in a robust way.
- 34
- Poisoning Attack Against Estimating From Pairwise Comparisons34
Experimental results show that the proposed methods can significantly reduce the performance of the ranker in the sense that the correlation between the true ranking list and the aggregated results with toxic data can be decreased dramatically.
- Performing scalable lossy compression on pixel encrypted images34
The experimental results show that the proposed scheme achieves much better performance than the existing lossy compression scheme for pixel-value encrypted images and also similar performance as the state-of-the-art lossy compressed for pixel permutation-based encrypted images.
- R1-Compress: Long Chain-of-Thought Compression via Chunk Compression and Search33
This work proposes R1-Compress, a two-stage chunk-level compression framework that segments Long-CoT into manageable chunks, applies LLM-driven inner-chunk compression, and employs an inter-chunk search mechanism to select the short and coherent sequence.
- 33
- Optimizing Two-Way Partial AUC With an End-to-End Framework33
A generic framework to construct surrogate optimization problems, which supports efficient end-to-end training with deep learning, is proposed and theoretical analyses show that the objective function of the surrogate problems will achieve an upper bound of the original problem under mild conditions.
- 33
- A Unified Perspective for Loss-Oriented Imbalanced Learning via Localization32
A principled learning algorithm is developed based on localized calibration to provide consistency validation across a broader range of losses and localized Lipschitz continuity to provide a fine-grained generalization bound.
- 32
- A Principled Design of Image Representation: Towards Forensic Tasks32
A new representation framework for forensics is proposed, called dense invariant representation (DIR), which is characterized by stable description with mathematical guarantees, and initially explored in passive and active forensics, exhibiting the benefits in fulfilling the requirements of such forensic tasks.
- Quaternion-Based Knowledge Graph Network for Recommendation32
This paper proposes Quaternion-based Knowledge Graph Network (QKGN) for recommendation, which represents users and items with quaternion embeddings in hypercomplex space, so that the latent inter-dependencies between entities and relations could be captured effectively.
- 32
- 32
- Uncertainty-Aware Medical Diagnostic Phrase Identification and Grounding31
This study represents a pioneering exploration of the MRG task, marking the first-ever endeavor in this domain and demonstrating the applicability of uMedGround in medical visual question answering and class-based localization tasks, where it highlights visual evidence aligned with key diagnostic phrases.
- CLIPVQA: Video Quality Assessment via CLIP30
An efficient and effective CLIP-based Transformer method for the VQA problem (CLIPVQA), which achieves new state-of-the-art VQA performance and up to 37% better generalizability than existing benchmark VQA methods.
- 30
- Interpreting Object-level Foundation Models via Visual Precision Search29
This paper bypasses internal model parameters to overcome attribution issues from multimodal fusion, dividing inputs into sparse sub-regions and using consistency and collaboration scores to accurately identify critical decision-making regions, and proposes a Visual Precision Search method that generates accurate attribution maps with fewer regions.
- Ada-R1: Hybrid-CoT via Bi-Level Adaptive Reasoning Optimization29
This work constructs a hybrid reasoning model by merging long and short CoT models to enable diverse reasoning styles, and applies bi-level preference training to guide the model to select suitable reasoning styles (group-level), and prefer concise and correct reasoning within each style group (instance-level).
- Efficient and Effective Weight-Ensembling Mixture of Experts for Multi-Task Model Merging29
An efficient-and-effective WEMoE (E-WEMoE) method, whose core mechanism involves eliminating non-essential elements in the critical modules of WEMoE and implementing shared routing across multiple MoE modules, thereby significantly reducing both the trainable parameters, the overall parameter count, and computational overhead of the merged model by WEMoE.
- Fine-Grained Multimodal DeepFake Classification via Heterogeneous Graphs29
A positional coding-based heterogeneous graph construction method that converts an audio-visual sample into a multimodal heterogeneous graph according to relevant hyperparameters and a cross-modal graph interaction module is devised to utilize audio-visual synchronization patterns for capturing inter-modal complementary information.
- 29
- 29
- Audio Visual Attribute Discovery for Fine-Grained Object Recognition29
This paper introduces a novel feature named audio visual attributes via discovering the correlations between the visual and audio representations and proposes a unified framework for training with video-level category label which can be implemented end-to-end in the step of inference.
- Halftone Image Watermarking by Content Aware Double-Sided Embedding Error Diffusion29
A performance analysis from a probabilistic perspective to introduce the error diffusion-based halftone visual watermarking (EDHVW) methods’ expected performances and limitations and proposes a new general EDHVW method, content aware double-sided embedding error diffusion (CaDEED), via considering the expected watermark decoding performance.
- A Hierarchical Distributed Processing Framework for Big Image Data29
Representative results of comprehensive experiments on the challenging ImageNet dataset are selected to validate the capacity of the proposed ICP framework over the traditional state-of-the-art methods, both in time efficiency and quality of results.
- 29
- 28
- 28
- DI-Retinex: Digital-Imaging Retinex Model for Low-Light Image Enhancement27
This paper proposes a new expression called Digital-Imaging Retinex model (DI-Retinex), which includes an offset term in the enhancement model, which allows for pixel-wise brightness contrast adjustment with a non-linear mapping function and design a variance suppression loss for regulating the additional offset term.
- Semantic-Aligned Adversarial Evolution Triangle for High-Transferability Vision-Language Attack27
This work proposes to generate AEs in the semantic image-text feature contrast space, which can project the original feature space into a semantic corpus subspace that can reduce the image feature redundancy, thereby improving adversarial transferability.
- Coarse-to-Fine Proposal Refinement Framework for Audio Temporal Forgery Detection and Localization27
A novel coarse-to-fine proposal refinement framework that incorporates a frame-level detection network (FDN) and a proposal refinement network (PRN) for audio temporal forgery detection and localization and designs a boundary-aware feature enhancement module to capture the contextual information of multiple transition boundaries and guide the interaction between boundary information and temporal features via a cross-attention mechanism.
- Object Detectors in the Open Environment: Challenges, Solutions, and Outlook27
This paper aims to provide a fresh, comprehensive, and systematic understanding of the challenges and solutions associated with open-environment object detectors, thus catalyzing the development of more solid applications in real-world scenarios.
- A Tale of HodgeRank and Spectral Method: Target Attack Against Rank Aggregation Is the Fixed Point of Adversarial Game27
The effectiveness of the suggested target attack strategies is demonstrated by a series of toy simulations and several real-world data experiments, showing that the proposed methods could achieve the attacker's goal and the leading candidate of the perturbed ranking list is the designated one by the adversary.
- CUNet: A Compact Unsupervised Network for Image Classification27
A new weighted pooling method in which different weighting values of adjacent neurons are considered leads to improved classification since the network becomes more robust against small image distortions.
- Active link selection for efficient semi-supervised community detection27
By connecting uncertain nodes to their community hubs and disconnecting the inter-community edges, one can sharpen the block structure of adjacency matrix more efficiently than randomly labeling links as the existing methods did.
- Single view compositing with shadows27
This approach efficiently utilizes the geometric and photometric constraints extracted from a single target image to synthesize the shadows consistent with the overall target scene for the inserted objects.
- 26
- Exploring Inconsistent Knowledge Distillation for Object Detection with Data Augmentation26
This work proposes inconsistent knowledge distillation (IKD), a method which outperforms state-of-the-art KD baselines on one-stage, two-stage and anchor-free object detectors (at most +1.0 mAP).
- 26
- Size-invariance Matters: Rethinking Metrics and Losses for Imbalanced Multi-object Salient Object Detection25
This paper proposes a generic approach that evaluates each salient object separately and then combines the results, effectively alleviating the imbalance in SOD and develops an optimization framework tailored to this goal, achieving considerable improvements in detecting objects of different sizes.
- 25
- JANE: Jointly Adversarial Network Embedding25
A novel Joint Adversarial Network Embedding (JANE) framework is proposed to jointly distinguish the real and fake combinations of the embeddings, topology information and node features, which can be optimized via alternating stochastic gradient.
- 25
- Toward Unsupervised Graph Neural Network: Interactive Clustering and Embedding via Optimal Transport25
- 25
- 25
- 24
- Content-Aware Quantization Index Modulation: Leveraging Data Statistics for Enhanced Image Watermarking24
Two content-aware quantization index modulation (QIM) algorithms are proposed: CA-QIM and CAMD-QIM, which aim to improve the embedding distortion of QIM-based watermarking schemes by considering the statistics of the cover signal vectors and messages.
- Shape-Preserving Object Depth Control for Stereoscopic Images24
Experimental results show that the proposed shape-preserving object depth control method can control object depth and preserve the shape of the object effectively without sensible background distortion.
- UncTrack: Reliable Visual Object Tracking With Uncertainty-Aware Prototype Memory Network23
UncTrack is a novel uncertainty-aware transformer-based tracker that predicts the target localization uncertainty and incorporates this uncertainty information for accurate target state inference and outperforms the state-of-the-art methods.
- PSNEA: Pseudo-Siamese Network for Entity Alignment between Multi-modal Knowledge Graphs23
A Pseudo-Siamese Network for multi-modal Entity Alignment (PSNEA), the first model that tackles graph heterogeneity and scarcity of initial seeds in one unified framework, is proposed.
- Deep Self-Supervised t-SNE for Multi-modal Subspace Clustering23
A deep Self-supervised t-SNE method (StSNE) is proposed, which learns soft label features by multi-modal encoders and utilizes the common label feature to supervise soft label feature of each modal by adversarial training and reconstruction networks.
- 23
- 22
- Graph Contrastive Learning Reimagined: Exploring Universality22
A novel GCL framework, named gRaph cOntraStive Exploring uNiversality (ROSEN), designed to achieve this objective, which equips a local graph structure inference module, utilizing the Block Diagonal Property (BDP) of the affinity matrix extracted from node ego networks.
- SDGAN: Disentangling Semantic Manipulation for Facial Attribute Editing22
The proposed Semantic Disentangled GAN introduces two key concepts: a semantic disentanglement generator that assigns facial representations to distinct attribute-specific editing modules, enabling the decoupling of the facial attribute editing process, and a semantic mask alignment strategy that confines attribute editing to appropriate regions, thereby avoiding undesired modifications.
- 22
- Watermarking Based on Compressive Sensing for Digital Speech Detection and Recovery †22
A novel imperceptible, fragile and blind watermark scheme is proposed for speech tampering detection and self-recovery and a deep learning-based enhancement method is proposed and implemented to increase the SNR of recovered speech signal.
- Lagrangian relaxation graph matching22
A new effective graph matching relaxation method, called Lagrangian relaxation graph matching (LRGM), which aims to generate a relaxation model by incorporating the affine mapping constraint into the matching objective optimization.
- Unsupervised pixel-level video foreground object segmentation via shortest path algorithm22
This paper proposes an object-level method to extract the foreground object in the video, which does not only guarantee the continuity of segmentation result, but also works well even under the cases of fast motion and occlusion.
- 22
- 22
- Video Understanding and Content-Based Retrieval22
This year, the joint team of UCF and the University of Modena has participated in the following tasks: (1) shot bound-ary detection, (2) low-level feature extraction, (3) high-levelfeature extraction,(4) topic search and (5) BBC rushes man-agement.
- DAShip: A Large-Scale Annotated Dataset for Ship Detection Using Distributed Acoustic Sensing Technique21
In this proposed framework, YOLO models, primarily trained on DAShip, are used as ship detectors and ship feature classifiers, achieving accurate dark ship detection combined with AIS message and demonstrating competitive performance in ship feature classification.
- Omnidirectional Image Super-resolution via Bi-projection Fusion21
A novel Bi-Projection Omnidirectional Image Super-Resolution (BPOSR) network to take advantage of the geometric properties of the above two projections to achieve state-of-the-art performance on omnidirectional image super-resolution.
- 21
- Revisiting and Exploring Efficient Fast Adversarial Training via LAW: Lipschitz Regularization and Auto Weight Averaging21
An effective FGSM-based fast adversarial training method equipped with Lipschitz regularization and Auto Weight averaging, abbreviated as FGSM-LAW is proposed, demonstrating the superiority of this method over state-of-the-art fast adversarial training methods and the advanced standard adversarial training methods.
- Sim2Word: Explaining Similarity with Representative Attribute Words via Counterfactual Explanations21
A new interpretation method is proposed that explains the image similarity models by salience maps and attribute words and can be applied to evidential learning cases, such as finding the most characteristic attributes in a set of face images, and is verified on the VGGFace2 dataset.
- 21
- Nonpharmaceutical interventions contribute to the control of COVID-19 in China based on a pairwise model21
A pairwise epidemic model with NPIs is devised to analyze COVID-19 outbreak in China by using confirmed cases during February 3rd–22nd, 2020 and indicated by simulation that the traced high-risk contacts from incubated to susceptible decrease under NPIs, regardless of infected cases.
- Stochastic Non-Convex Ordinal Embedding With Stabilized Barzilai-Borwein Step Size21
This work proposes a stochastic algorithm called SVRG-SBB, which has the following features: SVD-free via dropping convexity, with good scalability by the use of stochastically reduced gradient, and adaptive step size choice via introducing a new stabilized Barzilai-Borwein (SBB) method.
- Beauty eMakeup21
A Beauty eMakeup System to automatically recommend the most suitable makeup for a female and synthesis the makeup on her face is presented and extensive experimental evaluations and analysis on testing images well demonstrate the effectiveness of the proposed system.
- 21
- Enhanced invasiveness in multidrug resistant leukemic cells is associated with overexpression of P-glycoprotein and cellular inhibitor of apoptosis protein21
It is shown that overexpression of P-gp and cIAP may enhance the infiltration of leukemic cells and co-localize with protein kinase C-ε (PKC-ε) in MDR variants.
- Focal-SAM: Focal Sharpness-Aware Minimization for Long-Tailed Classification20
Focal-SAM is introduced, which assigns different penalties to class-wise sharpness, achieving fine-grained control without extra backpropagations, thus maintaining efficiency and theoretically analyzing Focal-SAM's generalization ability and derive a sharper generalization bound.
- VL-Trojan: Multimodal Instruction Backdoor Attacks against Autoregressive Visual Language Models20
This work introduces a multimodal instruction backdoor attack, namely VL-Trojan, that facilitates image trigger learning through active reshaping of poisoned features and enhances black-box attack efficacy through an iterative character-level text trigger generation method.
- Fast Propagation Is Better: Accelerating Single-Step Adversarial Training via Sampling Subnetworks20
This work proposes to exploit the interior building blocks of the model to improve efficiency and achieves better model robustness by dynamically sample lightweight subnetworks as a surrogate model during training.
- Adaptive 3D Mesh Steganography Based on Feature-Preserving Distortion20
This work proposes a highly adaptive embedding algorithm, guided by the principle of minimizing a carefully crafted distortion through efficient steganography codes, that can preserve mesh features to a considerable extent, including steganalytic and geometric features, while achieving a high embedding capacity.
- FedZKP: Federated Model Ownership Verification with Zero-knowledge Proof20
It is shown that the FedZKP scheme without disclosing credentials is guaranteed to defeat a variety of existing and potential attacks.
- 20
- 20
- Deep Video Dehazing20
This work develops a deep learning solution for video dehazing, where a CNN is trained end-to-end to learn how to accumulate information across frames for transmission estimation and is subsequently used to recover a haze-free frame via atmospheric scattering model.
- 20
- One Image is Worth a Thousand Words: A Usability Preservable Text-Image Collaborative Erasing Framework19
This work proposes a novel solution by directly integrating visual supervision into the erasure process, introducing the first text-image Collaborative Concept Erasing (Co-Erasing) framework, which outperforms state-of-the-art erasure approaches significantly with a better trade-off between efficacy and usability.
- Generalized Semantic Contrastive Learning via Embedding Side Information for Few-Shot Object Detection19
This work proposes a generalized feature learning method for FSOD by leveraging side information, and introduces a side-information guided region-aware masking module that improves sample diversity by removing biased information through counterfactual explanations.
- 19
- CAT: Enhancing Multimodal Large Language Model to Answer Questions in Dynamic Audio-Visual Scenarios19
- 19
- Improved Diversity-Promoting Collaborative Metric Learning for Recommendation19
The fundamental limitation of the widely adopted hard-aware sampling from the One-Way Partial AUC (OPAUC) perspective is revealed and an effective sampling alternative for the CML-based paradigm is developed.
- 19
- 19
- 19
- DirMixE: Harnessing Test Agnostic Long-Tail Recognition With Hierarchical Label Vartiations18
A general Latent Skill Finetuning (LSF) framework for parameter-efficient finetuning of foundation models is developed and it is proved that the covering number of the PEFT hypothesis class scales with the number of trainable parameters.
- RobustSplat: Decoupling Densification and Dynamics for Transient-Free 3DGS18
This work introduces a delayed Gaussian growth strategy that prioritizes optimizing static scene structure before allowing Gaussian splitting/cloning, mitigating overfitting to transient objects in early optimization and designs a scale-cascaded mask bootstrapping approach that first leverages lower-resolution feature similarity supervision for reliable initial transient mask estimation, and then progresses to high-resolution supervision to achieve more precise mask prediction.
- Semi-Supervised State-Space Model with Dynamic Stacking Filter for Real-World Video Deraining18
A dual-branch spatiotemporal state-space model is proposed to enhance rain streak removal in video sequences and develops a median stacking loss to enable semi-supervised learning by generating pseudo-clean patches based on the sparsity prior of rain.
- 18
- 18
- 18
- Analysis of Potential Risk of COVID-19 Infections in China Based on a Pairwise Epidemic Model18
A pairwise epidemic model is devised to analyze the COVID-19 outbreak in China based on confirmed cases reported during the period February 3rd--17th, 2020 and gives an answer to the public concern on how long the stringent containment strategies should maintain.
- 18
- 18
- 18
- 18
- View invariant action recognition using weighted fundamental ratios18
This paper fully investigates the concept of fundamental ratios, demonstrates their application and significance in view-invariant action recognition, and proposes a generic method for weighting body point line segments and investigates the importance of different body parts in action recognition.
- Video synchronization and its application to object transfer18
This work uses fundamental ratios to synchronize video sequences of distinct scenes captured by cameras undergoing similar motions and presents a simple method to align 3D camera trajectories when the fundamental ratios are not able to match the noisy trajectories.
- SafeMobile: Chain-level Jailbreak Detection and Automated Evaluation for Multimodal Mobile Agents17
This work explores the security issues surrounding mobile multimodal agents, attempts to construct a risk discrimination mechanism by incorporating behavioral sequence information, and designs an automated assisted assessment scheme based on a large language model.
- 17
- Sequential Manipulation Against Rank Aggregation: Theory and Algorithm17
The game-theoretic and empirical evidence shows that the proposed method manipulates the results of rank aggregation methods in a sequential manner, and a distributionally robust estimator, which replaces the maximum likelihood estimation in a saddle point problem, provides a conservative data generation solution.
- 17
- 17
- 17
- 17
- 17
- 17
- Watch fashion shows to tell clothing attributes17
A novel semi-supervised method to predict clothing attributes with the assistance of unlabeled data like fashion shows is proposed, i.e., the unsupervised triplet network pre-training stage that ensures frames in the same video having coherent representations while frames from different videos having larger feature distances.
- 17
- 17
- Action recognition using 3D DAISY descriptor17
A recent image local descriptor, DAISY, is extended to three dimensions to deal with the information in the additional temporal domain in videos to create a novel spatial-temporal descriptor for action recognition.
- Video object segmentation with shortest path17
An object-level method to segment foreground object, while existing methods are normally based on low level information is proposed, which not only guarantees the continuity of segmentation result but also works well even under the big disturbance of fast motion object in the background.
- 17
- 16
- Task-Feature Collaborative Learning with Application to Personalized Attribute Prediction16
This work proposes a novel multi-task learning method called Task-Feature Collaborative Learning (TFCL), leveraging the collaborative grouping of features and tasks while suppressing inter-group knowledge sharing, and proposes an optimization method for TFCL with a global convergence guarantee.
- 16
- Adversarial Preference Learning with Pairwise Comparisons16
This work presents a novel adversarial framework for collaborative ranking that combines a deep generator to approximate an arbitrary continuous score function in terms of pairwise comparison and a discriminator that provides personalized supervision signals with increasing difficulty.
- Multi-Facet Network Embedding: Beyond the General Solution of Detection and Representation16
A novel network embedding method, Multi-facet Network Embedding (MNE), is proposed to capture the multiple facets of the network to improve the general solution for better modeling the real network.
- The (un)supervised NMF methods for discovering overlapping communities as well as hubs and outliers in networks16
This work defines a novel generative model and formalizes the detection of overlapping communities as well as hubs and outliers as an optimization problem on it and presents a community detection method, which detects all three structures altogether without prior knowledge of the number of communities.
- 16
- Motion matters16
This paper introduces a novel framework for compressing surveillance videos that significantly reduces the size of videos while gains much higher PSNR compared to the state of the art codecs.
- Duplication Localization and Segmentation16
This paper proposes a novel method to automatically detect and segment the duplicate regions within an image using the modified Efficient Subwindow Search algorithm, and generalizes duplication from traditional pure copy-paste to more general cases, which involves planar homography transformation.
- Detecting image forgeries using metrology16
This paper presents a framework for detecting fake regions using single view metrology and enforcing geometric constraints from shadows and shows that image forgery on the vertical plane or arbitrary plane can be detected through the measurement on such plane.
- 16
- Metrology from Vertical Objects16
This paper shows that the length ratio of two objects perpendicular to the reference plane can be expressed as a function of the camera principal point, and recovers the principal point and the camera focal length by minimizing the symmetric transfer error of geometric distances.
- 15
- 15
- Fast Ultra High-Definition Video Deblurring via Multi-scale Separable Network15
A novel deep model for fast and accurate UHD video deblurring (UHDVD) is proposed by a depth-wise separable-patch architecture, which operates with a multi-scale integration scheme to achieve a large receptive field without adding the number of generic convolutional layers and kernels.
- From Social to Individuals: A Parsimonious Path of Multi-Level Models for Crowdsourced Preference Aggregation15
A parsimonious mixed-effects model is proposed, which takes into account both the fixed effect that the majority of annotators follows a common linear utility model, and the random effect that some annotators might deviate from the common significantly and exhibit strongly personalized preferences.
- Improving the Efficiency and Effectiveness of Community Detection via Prior-Induced Equivalent Super-Network15
This work proposes a general, high-speed, effective and parameter-free semi-supervised community detection framework that transforms the original network into an equivalent but much smaller Super-Network, which perfectly ensures the must-link constraints and effectively encodes cannot- link constraints.
- 15
- A Stochastic Model for Detecting Overlapping and Hierarchical Community Structure15
This work proposes a generative model that employs a nonnegative matrix factorization formulization with a l2,1 norm regularization term, balanced by a resolution parameter, and derives the multiplicative update rule to learn the model parameters, and offers the proof of its correctness.
- Multi-cue Augmented Face Clustering15
A new algorithm, Multi-cue Augmented Face Clustering (McAFC), which effectively incorporates the cues via graph-guided sparse subspace clustering technique, and demonstrates the improvements of the algorithm over the state-of-the-art methods.
- MIFT: A framework for feature descriptors to be mirror reflection invariant15
This paper presents a framework, mirror-reflection invariant feature transform (MIFT), for improving their degenerated performance due to mirror reflections, and demonstrates the robust performance of MIFT.
- 15
- 15
- 14
- Decoupled Graph Energy-based Model for Node Out-of-Distribution Detection on Heterophilic Graphs14
DeGEM is introduced, which decomposes the learning process into two parts: a graph encoder that leverages topology information for node representations and an energy head that operates in latent space that surpasses previous state-of-the-art methods.
- 14
- 14
- 14
- Entity-Graph Enhanced Cross-Modal Pretraining for Instance-level Product Retrieval14
A novel Entity-Graph Enhanced Cross-Modal Pretraining (EGE-CMP) model is proposed for instance-level commodity retrieval, which explicitly injects entity knowledge in both node-based and subgraph-based ways into the multi-modal networks via a self-supervised hybrid-stream transformer.
- Self-Supervised Graph Neural Networks via Diverse and Interactive Message Passing14
A novel Diverse and Interactive Message Passing (DIMP) is proposed for self-supervised learning by overcoming over-smoothing issue and is equivalent to performing an overlapping community detection via expectation-maximization (EM).
- Duet Robust Deep Subspace Clustering14
A new robust deep framework called Duet Robust Deep Subspace Clustering (DRDSC) is contributed, to explicitly model the corrupted patterns from both the data reconstruction perspective and the latent self-expression perspective with two regularization norms.
- A Simple Framework for Face Photo‐Sketch Synthesis14
A simple framework for face photo-sketch synthesis that demands non feature localization, training or iteration process, creating vivid hair in sketch synthesis, and process arbitrary lighting conditions of input images, especially for complex self-shadows.
- 14
- 14
- 13
- Adversarial Backdoor Defense in CLIP13
Adversarial Backdoor Defense (ABD), a novel data augmentation strategy that aligns features with meticulously crafted adversarial examples, provides robust defense against both traditional uni-modal and multimodal backdoor attacks targeting CLIP.
- DI-Retinex: Digital-Imaging Retinex Theory for Low-Light Image Enhancement13
A new expression called Digital-Imaging Retinex theory (DI-Retinex) is proposed through theoretical and experimental analysis of Retinex theory in digital imaging, which includes an offset term in the enhancement model, which allows for pixel-wise brightness contrast adjustment with a non-linear mapping function.
- 13
- 13
- 13
- 12
- 12
- Hi-SIGIR: Hierachical Semantic-Guided Image-to-image Retrieval via Scene Graph12
This work proposes a novel hierarchical semantic-guided image-to-image retrieval method via scene graph, called Hi-SIGIR, which extracts and learns both the visual and semantic features of the nodes and relations within the scene graphs and introduces a surrogate that calculates relevance in a cross-modal manner to understand image content better.
- Propagation is All You Need: A New Framework for Representation Learning and Classifier Training on Graphs12
A novel GNN-specific training framework is proposed by simultaneously updating node representations and classifier parameters via a unified feature propagation scheme, which demonstrates the superior performance and the universality of the proposed framework to existing GNNs.
- TranSegPGD: Improving Transferability of Adversarial Examples on Semantic Segmentation12
An effective two-stage adversarial attack strategy to improve the transferability of adversarial examples on semantic segmentation, dubbed TranSegPGD, which can achieve state-of-the-art performance.
- FedSOV: Federated Model Secure Ownership Verification with Unforgeable Signature12
Experimental results on computer vision and natural language processing tasks demonstrate that FedSOV is an effective federated model ownership verification scheme enhanced with provable cryptographic security.
- Optimizing Partial Area Under the Top-K Curve: Theory and Practice12
Theoretical analysis shows that AUTKC has a better discrimination ability, and its Bayes optimal score function could give a correct top-<inline-formula><tex-math notation="LaTeX">$K$</tex-Math><alternatives><mml:math><mmL:mi>k</ mml:mi></mml?:math><inline-graphic xlink:href="wang-ieq4-3199970.gif"/></alternatives
- 12
- Collaborative Preference Embedding against Sparse Labels12
This work proposes a novel method named as Collaborative Preference Embedding (CPE) which directly deals with sparse and insufficient user preference information and leverages a compact embedding space by reducing the dependency across different dimensions of a code (embedding).
- 12
- AUTOMATIC GEO-REGISTRATION FOR PORT SURVEILLANCE12
The computed horizon and apex provide a metric rectification that removes the affine distortions and reduces the searching space for geo-registration in the nearly feature-less maritime video feeds.
- 12
- 11
- A High-Performance and Lightweight Maritime Target Detection Algorithm11
The experimental results show that the HPMTD achieves lightweight and high-precision detection of marine targets with a nearly 10 percent mean average precision value improvement with nearly half the model size, compared with counterparts.
- Asymmetrically Decentralized Federated Learning11
Empirically, extensive experiments conducted on the MNIST, CIFAR10&100 datasets demonstrate the superior performance of the proposed DFedSGPSM algorithm compared to several existing SOTA optimizers in terms of generalization.
- Phrase Grounding-Based Style Transfer for Single-Domain Generalized Object Detection11
This paper proposes a novel phrase-grounding-based style transfer (PGST) approach for single-domain generalized object detection that significantly improves mean average precision (mAP) and outperforms or matches the performance of domain-adaptive object detection methods, which require target domain data for training, in several challenging scenarios.
- 11
- 11
- 11
- Graph Neural Networks without Propagation11
The Low-Rank GNNs are proposed, whose key component is the low-rank attribute matrix approximation in ego-network, whose graph topology is employed to construct the ego-networks instead of message propagation, which is sensitive to topology noises.
- 11
- 11
- 11
- 11
- 11
- 11
- 10
- FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models10
The proposed FaceShield is capable of determining the authenticity of faces, identifying types of spoofing attacks, providing reasoning for its judgments, and detecting attack areas, and significantly outperforms previous deep learning models and general MLLMs on four FAS tasks, i.e., coarse-grained classification, fine-grained classification, reasoning, and attack localization.
- Robust Anti-Backdoor Instruction Tuning in LVLMs10
A lightweight, certified-agnostic defense framework, Robust Instruction Tuning, that finetunes only adapter modules and text embedding layers under instruction tuning that reduces their attack success rate to nearly zero, with an increase in training cost of less than 15%.
- Agent4FaceForgery: Multi-Agent LLM Framework for Realistic Face Forgery Detection10
This work introduces Agent4FaceForgery to address two fundamental problems: how to capture the diverse intents and iterative processes of human forgery creation, and how to model the complex, often adversarial, text-image interactions that accompany forgeries in social media.
- Promises and perils of using Transformer-based models for SE research10
It is demonstrated that encoder-only models (CodeBERT) can outperform encoder-decoder models for general-purpose coding tasks, and showcase the capability of decoder-only models (CodeGPT) for certain generation tasks.
- Impact of heterogeneous vaccination on epidemic dynamics in metapopulation networks10
Numerical results suggest that reducing the population size of large patches and decreasing population mobility from low-degree patches to high-degree patches are effective strategies to suppress epidemic spreading on the whole metapopulation network.
- Unsupervised stereoscopic image retargeting via view synthesis and stereo cycle consistency losses10
An unsupervised stereoscopic image retargeting network (USIR-Net), which consists of a content consistency term and a disparity consistency term, is proposed to preserve the structure information and prevent binocular disparity inconsistency and achieves superior performance compared with state-of-the-art methods.
- What to Select: Pursuing Consistent Motion Segmentation from Multiple Geometric Models10
A novel geometric-model-fusion framework for motion segmentation that incorporates the structural information shared by affinity matrices to select those semantically consistent entries and a multiplicative decomposition scheme is adopted to ensure structural consistency among multiple affinities.
- Autonomous Semantic Community Detection via Adaptively Weighted Low-rank Approximation10
This work develops a novel NMF-based method to autonomously determine the number of semantic communities and community structure simultaneously, showing its superior performance over several competing methods on synthetic and large real-world social networks.
- Parsimonious Mixed-Effects HodgeRank for Crowdsourced Preference Aggregation10
A parsimonious mixed-effects model based on HodgeRank that establishes a dynamic path from the common utility to individual variations, with different levels of parsimony or sparsity on personalization, based on newly developed Linearized Bregman Algorithms with Inverse Scale Space method.
- Image Retrieval and Ranking via Consistently Reconstructing Multi-attribute Queries10
This work discovers the correlation among attributes via expanding the query representation, and imposing the group sparsity on representations to reduce the disturbance of noisy data.
- Geometric attack resistant image watermarking based on MSER10
Experimental results on standard benchmark demonstrate that the proposed blind content-based image watermarking scheme is robust to geometric distortions as well as common signal processing.
- Lip segmentation and tracking under MAP-MRF framework with unknown segment number10
This paper proposes a lip segmentation and tracking scheme, featuring the robust performance to the estimate of the number of segment clusters, and presents a rival penalized iterative algorithm capable of performing segment clusters and over-segmentation elimination simultaneously.
- DADM: Dual Alignment of Domain and Modality for Face Anti-Spoofing9
This work proposes an alignment module between modalities based on mutual information, which adaptively enhances favorable modalities while suppressing unfavorable ones, and employs a dual alignment optimization method that aligns both sub-domain hyperplanes and modality angle margins, thereby mitigating domain gaps.
- 9
- 9
- Spatial-Frequency Discriminability for Revealing Adversarial Perturbations9
A discriminative detector relying on a spatial-frequency Krawtchouk decomposition that demonstrates the uniqueness and usefulness of the detector, exhibiting competitive scores on several deep models and image sets against a variety of adversarial attacks.
- PS-Net: A Learning Strategy for Accurately Exposing the Professional Photoshop Inpainting9
Experimental results illustrate that PS-Net can successfully distinguish forged regions in elaborate inpainted images, outperforming several state-of-the-art solutions and is also robust against some postprocessing operations commonly used in Photoshop.
- Focusing on Flexible Masks: A Novel Framework for Panoptic Scene Graph Generation with Relation Constraints9
This work designs a proposal matching strategy, which replaces deterministic segmentation results with proposals extracted from the off-the-shelf panoptic head for label alignment, thereby ensuring the all-matching of training samples and presents an innovative concept of employing relation predictions to constrain segmentation.
- Difference Residual Graph Neural Networks9
Extensive experiments demonstrate the superiority of DRC on real networks with both homophily and heterophily, and show that DRC can automatically determine the model depth and be adaptive to both shallow and deep models with two complementary components.
- An Intrusion Detection System for the Internet of Things Based on the Ensemble of Unsupervised Techniques9
UTEN-IDS uses the ensemble of autoencoders to handle the network data and performs the anomaly detection by an Isolation Forest algorithm and proves its utility in the IoT network when compared to other approaches.
- 9
- 9
- 9
- Augmented Image Retrieval using Multi-order Object Layout with Attributes9
A novel image search method is proposed, which allows the users to indicate any number of Regions Of Interest (ROIs) within the query as well as utilize various semantic concepts and spatial relations to search images.
- Action Recognition Using Subtensor Constraint9
This work finds that the trifocal tensor resides in a twelve dimensional subspace of the original space if the first two views are already matched and the fundamental matrix between them is known, which is referred to as subtensor, and uses the subtensor to perform the task of action recognition under three views.
- 9
- 8
- ID-Guard: A Universal Framework for Combating Facial Manipulation via Breaking Identification8
A universal framework for combating facial manipulation, termed ID-Guard, and a novel Identity Destruction Module (IDM) to suppress identifiable features in manipulated faces, which can be seamlessly integrated as a plug-and-play component into other tasks, such as adversarial training.
- FaceInsight: A Multimodal Large Language Model for Face Perception8
FaceInsight is proposed, a versatile face perception MLLM that provides fine-grained information and introduces visual textual alignment of facial knowledge to model both uncertain dependencies and deterministic relationships among facial information, mitigating the limitations of language-driven reasoning.
- Graph Contrastive Learning with Joint Spectral Augmentation of Attribute and Topology8
A novel GCL framework with a Joint spectrAl augMentation, named GCL-JAM, which is proved to perturb more inter-class edges and noise attributes compared to separate augmentation methods and validate the effectiveness and universality of GCL-JAM.
- Representing Noisy Image Without Denoising8
This paper explores a non-learning paradigm that aims to derive robust representation directly from noisy images, without the denoising as pre-processing, and introduces fractional-order parameter that offers time-frequency analysis capability that is not available in classical methods.
- Optimizing Multispectral Object Detection: A Bag of Tricks and Comprehensive Benchmarks8
This work proposes the first fair and reproducible benchmark specifically designed to evaluate the training"techniques", which systematically classifies existing multispectral object detection methods, investigates their sensitivity to hyper-parameters, and standardizes the core configurations.
- 8
- A Unified Framework against Topology and Class Imbalance8
This work develops a unified topology-aware AUC optimization (TOPOAUC) framework, which could simultaneously deal with the topology and class imbalance problem in graph learning and proposes a T opology-A ware I mportance L earning mechanism (TAIL), which considers theTopology of pairwise nodes and different contributions of topology information to pairwise node neighbors.
- 8
- 8
- 8
- N-intertwined SIS epidemic model with Markovian switching8
It is proved that the model modeled by stochastic switched differential equations based on the intertwined SIS model using a continuous time finite-state Markov chain has a unique global positive solution and establishes sufficient conditions for stochastically extinction and permanence of the epidemic.
- Double-Spending Detection for Fast Bitcoin Payment Based on Artificial Immune8
Experimental result shows that this model, composed of many immune-based Bitcoin nodes which include a detection modula and a traditional node, can efficiently detect double-spending attacks in fast Bitcoin payment.
- Exploring the roles of cannot-link constraint in community detection via Multi-variance Mixed Gaussian Generative Model8
This is the first work on discovering and exploring the importance of cannot-link constraints in semi-supervised community detection and proposes to characterize this process as a Multi-variance Mixed Gaussian Generative (MMGG) Model to address diverse degrees of confidences that exist in network topology and pairwise constraints and formulate it as a weighted nonnegative matrix factorization problem.
- Hyperlink-Aware Object Retrieval8
This paper proposes an efficient object retrieval technique by hyperlinking the visual entities among the reference data set at subimage level, and proposes a scalable object mining technique using Thread-of-Features, which is designed for mining subimage-level objects.
- Quadratic projection based feature extraction with its application to biometric recognition8
Experimental results demonstrate the superiority of the DualQML-based feature extraction algorithm compared to the current state-of-the-art algorithms on biometric recognition tasks.
- A flexible framework of adaptive method selection for image saliency detection8
An instance-aware method selection approach which aims to select a single method instead of aggregating the results of all candidate ones, which is more efficient and flexible to employ metric learning technique.
- Embedded omni-vision navigator based on multi-object tracking8
This paper presents an embedded omni-vision navigation system which involves landmark recognition, multi-object tracking, and vehicle localization, and Experimental results demonstrated that the navigator can efficiently offer the vehicle guidance.
- 8
- 8
- GAHE: Geometry-Aware Embedding for Hyper-Relational Knowledge Graph Representation7
GAHE adopts a multi-curvature geometry-aware approach by modeling HKGs in Euclidean space, hyperbolic space, and hyperspherical space (positive curvature) in a unified framework and introduces a module termed hyper-relational subspace learning, which allocates multiple sub-relations for each hyper-relation.
- 7
- Critical Forgetting-Based Multi-Scale Disentanglement for Deepfake Detection7
A novel disentangled representation learning framework for deepfake detection that extends reconstruction loss from image-level to feature-level, and proposes a critical forgetting mechanism that forces the detector to forget the most salient features during training, which correspond to specific forgery artifact types in the training set.
- 7
- Dynamic Analysis and Adaptive Discriminator for Fake News Detection7
The Monte Carlo Tree Search algorithm is introduced to leverage the self-reflective capabilities of large language models (LLMs) for prompt optimization, providing richer, domain-specific details and guidance to the LLMs, while enabling more flexible integration of LLM comment on news content.
- 7
- GAUSS: GrAph-customized Universal Self-Supervised Learning7
Novel GrAph-customized Universal Self-Supervised Learning (GAUSS) is proposed by exploiting local attribute distribution to replace the global parameters with locally learnable propagation.
- 7
- Two-Tier Data Packing in RLWE-based Homomorphic Encryption for Secure Federated Learning7
This work presents a novel two-tier data encoding approach tailored for RLWE-based HE, effectively utilizing RLWE's polynomial structure and proposes a new subring decomposition method, i.e., employing ring isomorphism mappings to project multiple large integers into varied sub-polynomial rings.
- Efficient Backdoor Defense in Multimodal Contrastive Learning: A Token-Level Unlearning Method for Mitigating Threats7
This study proposes an efficient defense mechanism against backdoor threats using a concept known as machine unlearning, strategically creating a small set of poisoned samples to aid the model's rapid unlearning of backdoor vulnerabilities, known as Unlearn Backdoor Threats (UBT).
- Fast Graph Sharpness-Aware Minimization for Enhancing and Accelerating Few-Shot Node Classification7
A novel algorithm, Fast Graph Sharpness-Aware Minimization (FGSAM), that integrates the rapid training of Multi-Layer Perceptrons (MLPs) with the superior performance of GNNs and reutilizes the gradient from the perturbation phase to incorporate graph topology into the minimization process at almost zero additional cost.
- A 2 SC: Adversarial Attacks on Subspace Clustering7
This article formally addresses the unexplored adversarial attacks in the equally important unsupervised clustering field and proposes the concept of the adversarial set and adversarialSet attack for clustering.
- SIGMA-DF: Single-Side Guided Meta-Learning for Deepfake Detection7
A novel method named Single-sIde Guided Meta-leArning framework for DeepFake detection (SIGMA-DF) which simulates the cross-domain scenarios during training by synthesizing virtual testing domain through meta-learning, which achieves state-of-the-art performance on unseen Deepfake data.
- 7
- 7
- 7
- 7
- Deep Robust Subjective Visual Property Prediction in Crowdsourcing7
A deep SVP prediction model is constructed which not only leads to better detection of annotation outliers but also enables learning with extremely sparse annotations, and a comparison multi-graph is constructed based on the collected annotations, where different labeling results correspond to edges with different directions between two vertexes.
- 7
- Image aesthetics enhancement using composition-based saliency detection7
A novel saliency detection method is presented, which utilizes the knowledge of photographic composition as priors to improve theSaliency detection results, and an online parameter selection method is proposed when utilizing GrabCut to achieve the saliency segmentation result.
- Visual saliency detection based on photographic composition7
A novel saliency detection method is presented, which utilizes the knowledge of photographic composition as priors to improve theSaliency detection results and an online parameter selection method is proposed when utilizing GrabCut to achieve the saliency segmentation result.
- 7
- Vision Enhancing LLMs: Empowering Multimodal Knowledge Storage and Sharing in LLMs6
Modular Visual Memory (MVM), a component integrated into the internal blocks of LLMs, designed to store open-world visual information efficiently, is introduced, a component integrated into the internal blocks of LLMs designed to store open-world visual information efficiently.
- SUMI-IFL: An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints6
A novel information-theoretic IFL framework named SUMI-IFL is introduced that imposes sufficiency-view and minimality-view constraints on forgery feature representation to achieve an accurate and concise forgery feature representation that counters the interference of task-unrelated features.
- UMDATrack: Unified Multi-Domain Adaptive Tracking under Adverse Weather Conditions6
This paper proposes UMDATrack, which is capable of maintaining high-quality target state prediction under various adverse weather conditions within a unified domain adaptation framework and proposes a target-aware confidence alignment module (TCA) following optimal transport theorem.
- 6
- OPEN: Orthogonal Propagation with Ego-Network Modeling6
A novel Orthogonal Propagation with Ego-Network modeling (OPEN) is proposed as modeling high-order relationships beyond pairwise one, preventing overfitting, robustness, and high efficiency.
- Deep Partial Rank Aggregation for Personalized Attributes6
This paper proposes an end-to-end partial ranking model which consists of a deep backbone architecture and a probabilistic model that captures the generative process of the partial rank annotations and is equipped with an adaptive perception threshold.
- 6
- 6
- Robust Ordinal Embedding from Contaminated Relative Comparisons6
A unified framework to jointly identify the contaminated comparisons and derive reliable embeddings is proposed and instead of considering the nuclear norm heuristics, it adopts an exact solution for rank equality constraint.
- Learning Personalized Attribute Preference via Multi-Task AUC Optimization6
A novel multi-task method to understand and predict personalized attribute annotations and proposes a multi-level task parameter decomposition to capture the evolution from a highly popular opinion of the mass to highly personalized choices that are special for each person.
- From Common to Special: When Multi-Attribute Learning Meets Personalized Opinions6
In the proposed model, the diversity of personalized opinions and the intrinsic relationship among multiple attributes are unified in a common-to-special manner and the model integrates a common cognition factor, an attribute- specific bias factor and a user-specific bias factor.
- Who to Ask6
Stile, an end-to-end intelligent fashion consultant system, to generate stylish outfits for given items via a bidirectional LSTM that can guarantee that items in the same outfit should share a similar style and neither redundant nor missing items exist in the resulting outfit for essential categories.
- 6
- Human Skin Detection via Semantic Constraint6
Different from traditional skin detection algorithms, the semantic constraint is introduced, which is based on the dependence between skin pixels and human body parts (skin pixels should be overlapped with body parts) to refine the detection.
- Feature selection with spatial path coding for multimedia analysis6
A model of spatial path coding (SPC) that uses a supervised technique to select sparse features that shows better performance in terms of classification and recognition accuracy as compared with state-of-the-art algorithms is proposed.
- Blind Robust Watermarking Mechanism Based on Maxima Curvature of 3D Motion Data6
Experimental results show that the proposed watermarking scheme is robust against many possible attacks such as uniform affine transforms (scaling, rotation and translation), noise addition, reordering and cropping.
- Map-based Active Leader-Follower Surveillance System6
A system to learn various PTZ camera characteristics which facilitate stable and accurate control of the camera during active tracking and evaluate the performance of each of the components individually.
- 6
- G 2 V 2 former: Graph Guided Video Vision Transformer for Face Anti-Spoofing5
The Graph Guided Video Vision Transformer (G2V2former) is proposed, which combines faces with facial landmarks for the fusion of photometric and dynamic features and designs a novel temporal attention, called Kronecker temporal attention, which has a wider receptive field and is beneficial for capturing dynamic information.
- No Query, No Access5
The hierarchical substitution model design is proposed, generating substitute models to mitigate the failure of a single substitute model at the decision boundary, and diverse adversarial example generation is used, employing various attack methods to generate and select the adversarial example with better similarity and attack effectiveness.
- 5
- 5
- 5
- Fair Outsourcing Paid in Fiat Money Using Blockchain5
A blockchain-based fair outsourcing scheme to support payment in fiat money is proposed, which has broad compatibility and, as an example, is subsequently instantiated with a conventional outsourcing solution of eigen-decomposition.
- Evaluating Visual Properties via Robust HodgeRank5
A robust ranking framework based on the principle of Hodge decomposition of imbalanced and incomplete ranking data is proposed that facilitates a novel outlier detection model as Huber’s LASSO in robust statistics.
- The Impact of Isolation on the Transmission of COVID-19 and Estimation of Potential Second Epidemic in China5
Simulations show that the isolation of close contacts tracked measure can efficiently contain the transmission of the potential second epidemic of COVID-19, and with isolation of all susceptible people or all infected people or both, there is no potential secondEpidemic of CO VID-19.
- Generalized Block-Diagonal Structure Pursuit: Learning Soft Latent Task Assignment against Negative Transfer5
A novel multi-task learning method, which simultaneously learns latent task representations and a block-diagonal Latent Task Assignment Matrix, and proposes an alternating optimization method which tells how negative transfer is alleviated in the method but also reveals an interesting connection between the method and the optimal transport problem.
- 5
- Magic pencil5
A novel image synthetic approach named Magic Pencil, which aims to synthesize the real images based on the sketch by designing a novel multi-task neural network based on generative adversarial nets composed of three components i.e. the generator, the discriminator, and the classifiers.
- 5
- The (un)supervised detection of overlapping communities as well as hubs and outliers via (bayesian) NMF5
A novel model to identify overlapping communities as well as hubs and outliers is defined and a normalized symmetric nonnegative matrix factorization algorithm is proposed to learn the parameters of the model.
- Classification using distances from samples to linear manifolds5
On most of the data sets, DSL outperforms the comparing classifiers based on k nearest samples or subspaces, and is even superior to support vector machines on some data sets.
- 5
- 5
- 4
- 4
- 4
- Top-K Pairwise Ranking: Bridging the Gap Among Ranking-Based Measures for Multi-label Classification4
This paper proposes a novel measure named Top-K Pairwise Ranking (TKPR), and a series of analyses show that TKPR is compatible with existing ranking-based measures and establishes an empirical surrogate risk minimization framework for TKPR.
- 4
- PTUS: Photo-Realistic Talking Upper-Body Synthesis via 3D-Aware Motion Decomposition Warping4
A Photo-realistic Talking Upper-body Synthesis method via 3D-aware motion decomposition warping, named PTUS, to both precisely synthesize the upper body as well as recover the details of the face such as blinking and lip synchronization, is proposed.
- Long Short-Term Graph Memory Against Class-imbalanced Over-smoothing4
A novel node-dependent residual connection is proposed to employ Long Short-Term Memory (LSTM), to model the sequence of node representation, i.e., the representations of one node in different layers can be seen as a sequence of states.
- 4
- Recurrent Meta-Learning against Generalized Cold-start Problem in CTR Prediction4
A recursive meta-learning model with the MAML (Model-Agnostic Meta-Learning)-like meta- learning method is proposed to make the model adapt to new tasks rapidly and a parallel structure for extracting the feature interactions to efficiently fuse attention mechanisms and the RNN layer is proposed.
- Semantic Correspondence with Geometric Structure Analysis4
The Flip-aware Distance Ratio method (FDR) is introduced to solve the correspondence problem for semantically similar images from the perspective of geometric structure analysis, and a Gradient Guided Simulated Annealing (GGSA) algorithm is proposed to perform a robust discrete optimization.
- Modeling the second outbreak of COVID-19 with isolation and contact tracing4
A mathematical model based on the epidemiology of COVID-19 is developed, incorporating the isolation of healthy people, confirmed cases and contact tracing measures, and it is verified that the current isolation method effectively contains the transmission of CO VID-19.
- 4
- Human parsing by weak structural label4
This paper proposes a weakly supervised deep parsing method to alleviate the human from the time-consuming labeling and develops a novel logistic pooling layer to constrain that the pixels responding to the color and corresponding category labels are the same to interpret the structural label.
- 4
- 4
- 4
- Image composite authentication using a single shadow observation4
This paper presents a novel method for image composite authentication based on a single shadow in an outdoor image, and is the first attempt, to the best knowledge, to make use of sun elevation with shadow of one object to authenticate outdoor images.
- TrojanTO: Action-Level Backdoor Attacks against Trajectory Optimization Models3
TrojanTO is proposed, the first action-level backdoor attack against Trajectory Optimization models and exhibits broad applicability to DT, GDT, and DC, underscoring its scalability across diverse TO model architectures.
- RobustSplat++: Decoupling Densification, Dynamics, and Illumination for In-the-Wild 3DGS3
This work introduces a delayed Gaussian growth strategy, a scale-cascaded mask bootstrapping approach that first leverages lower-resolution feature similarity supervision for reliable initial transient mask estimation, and then progresses to high-resolution supervision to achieve more precise mask prediction.
- 3
- 3
- Anti-Fake Vaccine: Safeguarding Privacy Against Face Swapping via Visual-Semantic Dual Degradation3
A novel privacy-preserving framework, named Anti-Fake Vaccine, to protect the facial images against the malicious face swapping, and develops an additive perturbation strategy to discover the shared adversarial perturbations across diverse face swapping models.
- 3
- 3
- 3
- 3
- A new encryption scheme for surveillance videos3
Experimental results demonstrate that the proposed method is robust to known cryptanalytic attacks, and enhances the overall security due to the foreground and background separation, and is faster than competing methods, which do not conduct foreground extraction.
- Erratum: Identifying overlapping communities as well as hubs and outliers via nonnegative matrix factorization3
A novel model where vertices are measured by their centrality in communities is proposed, and the identification of overlapping communities, hubs, and outliers is defined as an optimization problem, calculated by nonnegative matrix factorization.
- Unified Dictionary Learning and Region Tagging with Hierarchical Sparse Representation3
The proposed approach, called Unified Dictionary Learning and Region Tagging with Hierarchical Sparse Representation, has better performance of region tagging than the current state of the art methods.
- Object coding on the semantic graph for scene classification3
The experimental results show that the Object Coding on semantic graph can improve the classification accuracy and is proposed to address the scene classification problem.
- 3
- 3
- An Efficient Data Driven Algorithm for Multi-Sensor Alignment3
Qualitative and quantitative results are presented that show the ability of the proposed algorithm to handle large fraction of outliers in the data, spatial noise, and high traffic densities, which are defining characteristics of these applications.
- Multi-Source Information Driven Spatio-Temporal Hypergraph Learning for Traffic Forecasting2
MSTHG introduces a multi-source hypergraph fusion strategy that jointly models dynamic high-order spatial and temporal correlations and outperforms state-of-the-art baselines.
- 2
- LightSAM: Parameter-Agnostic Sharpness-Aware Minimization2
Theoretical results show that under weak assumptions, LightSAM could converge ideally with any choices of perturbation radius and learning rate, thus achieving parameter-agnostic in practical applications.
- DSPF: Dual-Stage Preservation and Fusion for Source-Free Domain Adaptive Point Cloud Completion2
This paper proposes a novel method called ''Dual-Stage Preservation and Fusion'' (DSPF), which comprises two key training stages tailored to this new setting: source-free domain adaptive point cloud completion and target fusion, and validates its effectiveness and robustness in source-free domain adaptation scenarios.
- 2
- 2
- 2
- Multi-task Adversarial Attacks against Black-box Model with Few-shot Queries2
CEMA simplifies complex multi-task scenarios by using a deep-level substitute model trained in aplug-and-play manner for text classification, enabling attacks without mimicking the victim model, converting multi-task attacks into classification attacks and allowing attacks across various tasks.
- Hierarchical Invariance for Robust and Interpretable Vision Tasks at Larger Scales2
This work shows how to construct over-complete invariants with a Convolutional Neural Networks (CNN)-like hierarchical architecture yet in a fully interpretable manner, and investigates the topic from theoretical, practical, and application perspectives.
- Boosting Semi-Supervised Learning with Dual-Threshold Screening and Similarity Learning2
A novel method to take full advantage of the unlabeled data, termed DTS-SimL, which includes two core designs: Dual-Threshold Screening and Similarity Learning and an overall loss function by assigning four different losses to four different types of data.
- 2
- Multi-Attention Based Visual-Semantic Interaction for Few-Shot Learning2
A novel Multi-Attention based Visual-Semantic Interaction (MAVSI) approach for FSL that utilizes spatial and channel attention mechanisms to effectively select discriminative visual features for the support set based on its ground-truth semantics while using all the support set semantics for each query set sample.
- 2
- Detecting Recolored Image by Spatial Correlation2
This paper tries to explore a solution from the perspective of the spatial correlation, which exhibits the generic detection capability for both conventional and deep learning-based recoloring and achieves the state-of-the-art detection accuracy on multiple benchmark datasets and exhibits well generalization for unknown types of recoloring methods.
- Updated Paired Regions for Shadow Detection from Single Image2
The physical principle of the shadowed region is analyzed and a paired region based algorithm for shadow detection without extra model training is presented and the proposed non-learning approach performs favorably against the state-of-the-art approaches.
- 2
- On the basic reproduction number in semi-Markov switching networks2
The basic reproduction number is the weighted average of the basic reproduction numbers of deterministic subnetworks that can determine epidemic persistence or extinction in theories and simulations.
- 3-in-1 Correlated Embedding via Adaptive Exploration of the Structure and Semantic Subspaces2
A novel generative model is proposed to formulate the generation process of the network and content from the embeddings, with respect to the Bayesian framework, and automatically leans to the information which is more discriminative.
- 2
- TPS_DR: A Universal Dimension Reducing Algorithm for Optimal Trust Path Selection in Complex Sensor Network2
A concept of complex trust network is introduced to model the relationships among network nodes in a complex sensor network and an algorithm TPS DR (Trust Path Selection based on Dimensionality Reduction) is proposed to simplify the trust network through reducing the dimensionality.
- 2
- 2
- Automatic Geo-registration and Inter-Sensor Calibration in Large Sensor Networks2
This chapter presents a data-driven approach to automatic and semi-automatic estimation of inter-sensor mapping, relative topology of overlapping cameras, and geo-registration in large visual sensor networks.
- 2
- 1
- Detecting Synthetic Image by Cross-Modal Commonality Interaction1
This work proposes a multi-modal high-frequency interactive detection framework for general synthetic image detection that explicitly integrates high-frequency information from both the spatial and frequency domains and achieves state-of-the-art performance in real-world detection scenarios.
- 1
- Beyond Progress Measures: Theoretical Insights into the Mechanism of Grokking1
This work investigates the grokking mechanism underlying the Transformer in the task of prime number operations and provides a unified perspective for understanding various previously proposed progress measures and introduces a novel, concise, and effective progress measure that could trace the changes in test loss more accurately.
- 1
- 1
- Object-Level Backdoor Attacks in RGB-T Semantic Segmentation with Cross-Modality Trigger Optimization1
This work is a pioneer of backdoor attacks in RGB-Thermal (RGB-T) semantic segmentation, overcome the critical limitation of current segmentation backdoor attacks that indiscriminately compromise all objects of a victim class and introduces a novel Object-level Backdoor Attack pipeline, termed OBA.
- 1
- 1
- PhaseWin Search Framework Enable Efficient Object-Level Interpretation1
A novel phase-window search algorithm that enables faithful region attribution with near-linear complexity and establishes a new state of the art in scalable, high-faithfulness attribution for object-level multimodal models.
- Towards Size-invariant Salient Object Detection: A Generic Evaluation and Optimization Approach1
A generic Size-Invariant Evaluation (SIEva) framework is proposed, which adheres to the size-invariant principle and significantly enhances the detection of salient objects across a broad range of sizes and is model-agnostic and can be seamlessly integrated with a wide range of SOD backbones.
- 1
- 1
- How Powerful Potential of Attention on Image Restoration?1
This paper proposes Continuous Scaling Attention (CSAttn), a method that computes attention continuously in three stages without using FFN and applies it to several image restoration tasks and shows that it can outperform CNN-based and Transformer-based image restoration approaches.
- 1
- 1
- Exploring the Robustness of Human Parsers Toward Common Corruptions1
This paper proposes a novel heterogeneous augmentation-enhanced mechanism to bolster robustness under commonly corrupted conditions and demonstrates good universality which can improve the robustness of the human parsing models and even the semantic segmentation models when facing various image common corruptions.
- Inter-frame Accelerate Attack against Video Interpolation Models1
A novel attack method named Inter-frame Accelerate Attack (IAA) is proposed that initializes the perturbation as the perturgation for the previous adjacent frame and reduces the number of attack iterations and can improve attack efficiency greatly.
- Confederated Learning: Going Beyond Centralization1
Systematic analysis shows that confederated learning generalizes traditional machine learning and the existing distributed paradigms like federation learning.
- 1
- 1
- 1
- 1
- A Vector Ionospheric Delay Correction Method for GNSS Positioning and Navigation Wide Area Augmentation Service1
A new ionospheric delay correction method, which uses the observation data of the ground monitoring stations to fit the triple, and then broadcast the triple to the user, to improve the ionsospheric correction accuracy of single frequency GNSS users.
- Who Likes What? — SplitLBI in Exploring Preferential Diversity of Ratings1
A multi-level model which learns both the common preference or utility function over the population based on features of alternatives to-be-compared, and preferential diversity functions conditioning on user categories is proposed.
- Less but Better: Generalization Enhancement of Ordinal Embedding via Distributional Margin1
A margin distribution learning paradigm for ordinal embedding, entitled Distributional Margin based Ordinal Embedding (DMOE), is proposed and a concise objective function is formulated which avoids maximizing margin mean and minimizing margin variance directly but exhibits the similar effect.
- 1
- 1
- LEAF1
A joint learning objective function which is composed of the high-order representation mining term and the classifier training term that can both preserve category-specific information and discover the common structure between categories.
- Research on Micro-blog New Word Recognition Based on MapReduce1
This paper first reduces noise to the corpus of micro-blog and employ the new filtering algorithm to filter the candidate words, then improves the traditional mutual information and adjacency entropy method respectively and put forward enhancement of mutual Information and relative adjacencies entropy.
- MatchDR1
A robust image correspondence technique from the perspective of spatial regularity is proposed, which addresses the visual deformation by introducing the spatial information by enforcing the distance ratio constrain and proposes a Gradient Guided Simulated Annealing method for robust optimization.
- 1
- Beautifying Fisheye Images using Orientation and Shape Cues1
A computational method for enhancing the aesthetics of fisheye images by exploiting the orientation and shape cues and can correct the rotation angle between the camera coordinate and the world coordinate to make the virtual camera oriented, and complete the missing part.
- 1
- Computing hypergraph Ramsey numbers by using quantum circuit1
A quantum algorithm to compute the two-color Ramsey numbers for uniform hypergraphs by using the quantum counting circuit is presented.
- 1
- HYPERGRAPH RAMSEY NUMBERS AND ADIABATIC QUANTUM ALGORITHM1
This paper considers that the two-color Ramsey numbers R(m, n; r) for r-uniform hypergraphs can be computed by using the similar ways in Phys.
- Photographic composite detection using circles1
This work proposes several new methods for detecting photographic composites using circles, focusing on three kinds of scenes: two coplanar circles with the same radius; a single circle with its discriminable center; and a singlecircle with geometric constraints for camera calibration.
- QUADRATIC TSALLIS ENTROPY BIAS AND GENERALIZED MAXIMUM ENTROPY MODELS1
The generalized quadratic Tsallis Entropy Bias Compensation (TEBC) Maxent can be expected to alleviate the overfitting and underfitting of the Maximum Entropy (Maxent) model.
- –
- –
- –
- –
- –
- –
- Toward Realistic Low-Light Image Enhancement via ISP-Driven Data Modeling–
This paper proposes an image signal processing (ISP)-driven data synthesis pipeline that addresses challenges by generating scalable paired training data and applies it to existing LLIE methods, demonstrating its impact on improving their practicality and generalizability in real-world applications.
- –
- –
- –
- Differential Privacy Enabled Cascaded Filter for Efficient and Privacy-Preserving Federated Learning–
- –
- –
- –
- –
- Closing the Approximation Gap of Partial AUC Optimization: A Tale of Two Formulations–
The approximation gap of PAUC optimization is closed by presenting two simple instance-wise minimax reformulations: one with an asymptotically vanishing gap, the other with the unbiasedness at the cost of more variables.
- PersGuard: Preventing Malicious Personalization in Text-to-Image Diffusion Models via Model Backdoors–
This paper introduces PersGuard, a novel backdoor-based approach that prevents malicious personalization of specific images, and significantly outperforms existing techniques, offering a more robust solution for privacy and copyright protection.
- –
- –
- –
- –
- –
- On the Mistaken Assumption of Interchangeable Deep Reinforcement Learning Implementations–
The results of studying the extent of implementation inconsistencies, their effect on the implementations' performance, as well as their impact on the conclusions of prior studies under the assumption of interchangeable implementations showed that they are not interchangeable.
- –
- –
- MMGIA: Gradient Inversion Attack Against Multimodal Federated Learning via Intermodal Correlation–
This work proposes MMGIA, an intermodal correlation-driven gradient inversion attack that systematically exploits multimodal correlation to enhance data reconstruction quality and reveals the heightened privacy risks in MMFL.
- Self-correcting text-aware strategy for text-caption–
This work improves the existing approach by proposing a text-aware recognizer to extract image text information from the input data and generate corresponding text descriptions and text features and introduces the caption-rectify module, which can better improve the text information involved in the image and model the text information recognized in the textcaps dataset.
- –
- ADT-SSL: Adaptive Dual-Threshold for Semi-Supervised Learning–
An Adaptive Dual-Threshold method for Semi-Supervised Learning (ADT-SSL), where except for the fixed threshold, ADT extracts another class-adaptive threshold from the labeled data to take full advantage of the unlabeled data whose predictions are less than 0.95 but more than the extracted one.
- –
- Real‐time target object tracking via dual‐directional scaling state predicting–
A novel dual-directional scaling state predicting method based on correlation filtering is proposed and the adaptive horizontal and vertical searches are adopted to separately and accurately estimate the scale changes in different directions.
- An Unsupervised Subspace Ranking Method for Continuous Emotions in Face Images.–
An unsupervised and novel approach for ranking of continuous emotions in images us-ing canonical polyadic decomposition and theoretical and empirical proof that this system is capable of generating a Pearson Correlation Coefficient that outperforms the state of the art by a large margin is provided.
- –
- –
- –
- –
- –
- –
- –
- Landmark measurement of human skull–
The landmark measurement of human skull is fundamental to geometric morphometry of palaeoanthropology and the proposed method can measure fast and accurately the distances between all pairs of landmarks and interior angles from triangulations of the landmarks.
- Multiple View Geometry For Video Analysis And Post-production–
The novel inter-image constraints are introduced that are particularly useful for scenes where minimal information is available and are used to develop an efficient framework for video analysis and post-production in multiple cameras.
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 2026-10-14. 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.