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
- Zhenya HeSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works382 from public data
- Learning Texture Transformer Network for Image Super-Resolution966
A novel Texture Transformer Network for Image Super-Resolution (TTSR), in which the LR and Ref images are formulated as queries and keys in a transformer, respectively, which achieves significant improvements over state-of-the-art approaches on both quantitative and qualitative evaluations.
- Generating Human Motion from Textual Descriptions with Discrete Representations798
A simple and must-known conditional generative framework based on Vector Quantised-Variational AutoEncoder (VQ-VAE) and Generative Pre-trained Transformer for human motion generation from textural descriptions shows better performance than competitive approaches, including recent diffusion-based approaches.
- Spatio-Temporal AutoEncoder for Video Anomaly Detection563
A novel model called Spatio-Temporal AutoEncoding (ST AutoEncoder or STAE), which utilizes deep neural networks to learn video representation automatically and extracts features from both spatial and temporal dimensions by performing 3-dimensional convolutions, which enhances the motion feature learning in videos.
- 364
- Deep convolutional neural network VGG-16 model for differential diagnosing of papillary thyroid carcinomas in cytological images: a pilot study288
After training with a large dataset, the DCNN VGG-16 model showed great potential in facilitating PTC diagnosis from cytological images and the contours, perimeter, area and mean of pixel intensity of PTC in fragmented images were more than the benign nodules.
- Natural Image Matting via Guided Contextual Attention198
This work develops a novel end-to-end approach for natural image matting with a guided contextual attention module, which is specifically designed for image matts and can mimic information flow of affinity-based methods and utilize rich features learned by deep neural networks simultaneously.
- 172
- Deep CTR Prediction in Display Advertising154
A novel deep neural network (DNN) based model is introduced that directly predicts the CTR of an image ad based on raw image pixels and other basic features in one step.
- 147
- Modified generalized projective synchronization of a new fractional-order hyperchaotic system and its application to secure communication144
A novel modified generalized projective synchronization (MGPS) is proposed based on the stability theory of the fractional-order system, where the states of the drive and response systems are asymptotically synchronized up to a desired scaling matrix.
- 143
- Adversarial and Contrastive Variational Autoencoder for Sequential Recommendation135
This work introduces the adversarial training for sequence generation under the Adversarial Variational Bayes (AVB) framework, which enables the model to generate high-quality latent variables and employs the contrastive loss.
- 134
- 123
- On stability of nonlinear continuous-time neural networks with delays120
The Lyapunov function method is utilized to analyze stability of continuous nonlinear neural networks with delays and obtain some new sufficient conditions ensuring the globally asymptotic stability independent of delays.
- Fragile watermarking scheme for image authentication111
A simple fragile watermarking scheme for image authentication is proposed that is secure, fast, and capable of detecting and localising modification, and is immune of the well-known quantisation attack.
- An Adversarial Approach to Hard Triplet Generation107
This work proposes an adversarial network for Hard Triplet Generation (HTG) to optimize the network ability in distinguishing similar examples of different categories as well as grouping varied examples of the same categories.
- Structural and Statistical Texture Knowledge Distillation for Semantic Segmentation92
This paper introduces a Contourlet Decomposition Module (CDM) that decomposes low-level features with iterative laplacian pyramid and directional filter bank to mine the structural texture knowledge and proposes a Denoised Texture Intensity Equalization Module (DTIEM) to adaptively extract and enhance statistical texture knowledge through heuristics iterative quantization and denoised operation.
- Digital image splicing detection based on approximate run length91
An approximate run length based scheme is proposed to detect image splicing and demonstrates that the proposed approach can achieve a relatively high accuracy with less computational cost and fewer features when compared with other methods.
- 91
- 89
- Using deep convolutional neural networks for multi-classification of thyroid tumor by histopathology: a large-scale pilot study88
The results indicate that DCNN models may have potential for facilitating histopathologic thyroid disease diagnosis, and especially VGG-19 achieved satisfactory accuracies on the task of differentiating thyroid tumors by histopathology.
- Hyperchaotic secure communication via generalized function projective synchronization87
Two different hyperchaotic secure communication schemes by using generalized function projective synchronization (GFPS), where the drive and response systems could be synchronized up to a desired scaling function matrix are presented.
- Deep learning based classification of ultrasound images for thyroid nodules: a large scale of pilot study85
Nodules that are 0.5-1.0 cm in size and have microcalcification and a taller shape can be more accurately diagnosed by Inception-v3, and can achieve an excellent diagnostic efficiency.
- Global exponential stability of delayed competitive neural networks with different time scales75
A time delay parameter is introduced into the neural network model to characterize the signal transmission delays in real neural systems and the finite switch speed in the circuit implementations of neural networks, and a new sufficient condition ensuring global exponential stability of the networks is derived.
- Show, Recall, and Tell: Image Captioning with Recall Mechanism74
Inspired by pointing mechanism in text summarization, this paper adopts a soft switch to balance the generated-word probabilities between SG and RWS in the CIDEr optimization step, and introduces an individual recalled-word reward (WR) to boost training.
- Enhancing human activity recognition using deep learning and time series augmented data73
This paper compares the performance of the Vanilla, Long-Short Term Memory, and Gated Recurrent Units neural network models on three open-source datasets and shows that using data augmentation significantly enhances recognition quality.
- Graph regularized discriminative non-negative matrix factorization for face recognition71
Both the graph Laplacian and supervised label information are jointly utilized to learn the projection matrix in the new model, and the corresponding multiplicative update solutions for the optimization framework are provided, together with the convergence proof.
- Deep convolutional neural network Inception-v3 model for differential diagnosing of lymph node in cytological images: a pilot study69
After training with a large dataset, the Inception-v3 DCNN model showed great potential in facilitating the diagnosis of cervical lymphadenopathy using cytological images and revealed that NHL was the most challenging cytology type for DCNN to differentiate.
- Enhanced modularity-based community detection by random walk network preprocessing69
This paper proposes to use a strategy to preprocess networks before feeding them into modularity-based algorithms, based on the observation that dynamic processes triggered on vertices in the same community possess similar behavior patterns but dissimilar on Vertices in different communities.
- A novel image watermarking scheme based on support vector regression69
A novel support vector regression based color image watermarking scheme is proposed that outperform the Kutter's method and Yu's method against different attacks including noise addition, shearing, luminance and contrast enhancement, distortion, etc.
- Cluster synchronization in the adaptive complex dynamical networks via a novel approach68
It is demonstrated that the synchronization performance is sensitively affected by the control gain, the inner-coupling matrix and the network topological structure.
- Generalized synchronization of the fractional-order chaos in weighted complex dynamical networks with nonidentical nodes66
This paper investigates generalized chaotic synchronization of the general fractional-order weighted complex dynamical networks with nonidentical nodes and finds that this effect sensitively depends on both the fractional order θ and the feedback gain ki.
- Robust watermarking based on DWT and nonnegative matrix factorization66
Because of the multiresolution decomposition for DWT and physically meaningful factorization for NMF, the proposed scheme can achieve good robustness, which is also demonstrated in the following experiments.
- Synchronization of chaotic neural networks via output or state coupling66
This work forms a general framework for the synchronization problem in which one chaotic neural network, working as the driving system, sends its output or state values to the other, which serves as the response system.
- Robust digital image watermarking based on subsampling65
Experimental results show good robustness and security of the proposed scheme and watermark extraction is simply by comparing the DCT coefficients of the watermarked subimages and does not need original image.
- 62
- 62
- 61
- Feature based robust watermarking using image normalization59
The proposed scheme combines the advantages of feature point detection and image normalization, which can achieve strong robustness to signal processing and geometrical distortions, and demonstrates good performance of the proposed scheme.
- Efficient linear discriminant analysis with locality preserving for face recognition58
A novel feature extraction algorithm which integrates both global and local geometric structures is proposed and experimental results show that the proposed integration framework is competitive with traditional face recognition algorithms, which use either global or local structure only.
- Outer synchronization of uncertain general complex delayed networks with adaptive coupling57
By means of the presented controllers and corresponding update laws, the outer synchronization between two complex delayed networks with adaptive coupling can be achieved without calculating the eigenvalues of the coupling configuration matrix.
- LRNNET: A Light-Weighted Network with Efficient Reduced Non-Local Operation for Real-Time Semantic Segmentation56
A light-weighted network with an efficient reduced non-local module (LRNNet) for efficient and realtime semantic segmentation and a factorized convolutional block in ResNet-Style encoder to achieve more lightweighted, efficient and powerful feature extraction.
- Non-negative and sparse spectral clustering54
It is proved that the spectral clustering under some conditions is equivalent to Non-negative matrix factorization (NMF), and under the NMF framework, several non-negative and sparse spectral clustersering algorithms are proposed.
- Generalized function projective (lag, anticipated and complete) synchronization between two different complex networks with nonidentical nodes54
Numerical simulations verify the effectiveness and feasibility of the proposed synchronization method and show that the synchronization rate is sensitively influenced by the feedback strength, the time delay, the network size and the network topological structure.
- Watermarking scheme based on support vector machine for colour images54
Owing to the good generalisation ability of the support vector machine, even when the watermarked image is heavily distorted the watermark can be successfully extracted.
- 53
- Ultra-High Resolution Segmentation with Ultra-Rich Context: A Novel Benchmark51
The URUR dataset is introduced, in the meaning of Ultra-High Resolution dataset with Ultra-Rich Context, and WSDNet, a more efficient and effective framework for UHR segmentation especially with ultra-rich context is proposed.
- 48
- Feature based watermarking using watermark template match48
Because the proposed feature detection is robust against JPEG compression, filtering, noise addition and geometric distortions, the proposed watermarking scheme can achieve good performance against these attacks, and experimental results also demonstrate the superiority.
- Prompt Guided Transformer for Multi-Task Dense Prediction46
This paper introduces a simple and lightweight task-conditional model called Prompt Guided Transformer (PGT), which incorporates task-specific prompts in the self-attention mechanism to achieve global dependency modeling and parameter-efficient feature adaptation across multiple tasks.
- A Fast Algorithm for Recovery of Jointly Sparse Vectors based on the Alternating Direction Methods46
A very fast algorithm is proposed, called MMV-ADM, to solve the jointly sparse signal recovery problem in MMV settings based on the alternating direction method (ADM), which is at least dozens of times faster than MMVprox with comparable recovery accuracy.
- 44
- A very fast algorithm for detecting community structures in complex networks44
A new algorithm is proposed, which uses only local information to analyze community structures in complex networks and demonstrates excellent detection results and very fast processing performance, much faster than the existing comparable algorithms of the same kind.
- Some sufficient conditions for global exponential stability of delayed Hopfield neural networks44
Some sufficient conditions for existence and uniqueness of equilibrium and global exponential stability in delayed Hopfield neural networks are derived by using a different approach from the usually used one where the existence, uniqueness ofilibrium and stability are proved in two separate steps.
- 43
- Locality Preserving Hashing42
A novel hashing algorithm called Locality Preserving Hashing is proposed, which learns a set of locality preserving projections with a joint optimization framework, which minimizes the average projection distance and quantization loss simultaneously.
- Finding communities in directed networks by PageRank random walk induced network embedding42
This work addresses the problem of finding communities in directed networks by using PageRank random walk induced network embedding to transform a directed network into an undirected one, where the information on edge directions is effectively incorporated into the edge weights.
- Global Exponential Stability of Multitime Scale Competitive Neural Networks With Nonsmooth Functions42
- Stability criteria for delayed neural networks42
Investigation of delay-independent global asymptotic and exponential stability for a class of delayed neural networks (DNN's) is investigated, and some criteria are established to ensure stability of DNN's by applying the Lyapunov direct method.
- 40
- A chaos-generator: analyses of complex dynamics of a cell equation in delayed cellular neural networks40
The presented model can serve as a chaos-generator, in which chaos can be generated from any one-dimensional (1-D) linear autonomous system just by the addition of a piecewise-linear delayed feedback.
- Classification of Motor Imagery EEG Signals with Deep Learning Models39
Recurrent neural networks and convolutional neural networks inspired by speech recognition and natural language processing are combined and deep models consist of stacking random forests to enhance the ability of feature representation and classification abilities for motor imagery EEG signals.
- Hybrid synchronization of the general delayed and non-delayed complex dynamical networks via pinning control39
This paper investigates the hybrid synchronization problem of two coupled complex dynamical networks with non-delayed and delayed coupling by the pinning control strategy using the LaSalle invariance principle and linear matrix inequality technique and obtains some sufficient conditions by applying simple linear feedback and adaptive controllers to a part of nodes.
- 39
- PA-SAM: Prompt Adapter SAM for High-Quality Image Segmentation36
This paper introduces a novel prompt-driven adapter into SAM, namely Prompt Adapter Segment Anything Model (PA-SAM), aiming to enhance the segmentation mask quality of the original SAM.
- Multi-modal constraint propagation for heterogeneous image clustering36
An effective solution to the multi-modal constraint propagation problem is provided by decomposing it into a set of independent multi-graph based two-class label propagation subproblems which are then merged into a unified problem and solved by quadratic optimization.
- 36
- PCA-based web page watermarking36
A novel watermarking scheme for the tamper-proof of web pages, which is free of this embarrassment and generates watermarks based on the principal component analysis (PCA) technique.
- 35
- Tree-of-Table: Unleashing the Power of LLMs for Enhanced Large-Scale Table Understanding34
Experiments across diverse datasets demonstrate that Tree-of-Table sets a new benchmark with superior performance, showcasing remarkable efficiency and generalization capabilities in large-scale table reasoning.
- Fedhca 2 : Towards Hetero-Client Federated Multi-Task Learning34
The Hyper Conflict-Averse Aggregation scheme is proposed, inspired by task interaction in MTL, which uses layer-wise cross attention to enhance decoder interactions while alleviating model incongruity and employs learnable Hyper Aggregation Weights for each client to customize personalized parameter updates.
- A General Common Spatial Patterns for EEG Analysis With Applications to Vigilance Detection34
A new efficient algorithm based on matrix deflation to solve the proposed NCSP algorithm and its extensions-nonparametric multi-class CSP (NMCSP) and Experimental results on EEG-based vigilance estimation and motor imagery recognition task demonstrate the effectiveness and efficiency of the proposed algorithms.
- 33
- 33
- Efficient Multi-Dimensional Tensor Sparse Coding Using t-Linear Combination31
This paper proposes two novel multi-dimensional tensor sparse coding schemes using the t-linear combination, based on the shifted versions of the bases, which outperform the state-of-the-art tensor-based sparse coding models.
- Novel robust image watermarking based on subsampling and DWT31
This paper presents a robust digital image watermarking scheme based on subsampling and DWT that achieves good robustness against JPEG compression, common image processing operation and geometric distortions.
- A PCA-based watermarking scheme for tamper-proof of web pages29
A novel watermarking scheme for tamper-proof of web pages that overwhelms existing methods of watermarked and Hash in that it does not increase the file size.
- Smart library book sorting application with intelligence computer vision technology27
A new image recognition method to identify books in libraries based on barcode decoding together with deep learning optical character recognition (OCR) and its application in library book identification processing is described.
- Pairwise constrained concept factorization for data representation26
This paper proposes a novel semi-supervised concept factorization method, called Pairwise Constrained Concept Factorization (PCCF), which incorporates pairwise constraints into the CF framework, and expects that data points which have pairwise must-link constraints should have the same class label as much as possible, while data points with pairwise cannot- link constraints will have different class labels.
- Semi-supervised non-negative matrix factorization for image clustering with graph Laplacian25
This paper proposes a novel semi-supervised non-negative matrix factorization (SEMINMF) algorithm, which not only utilizes the local structure of the data characterized by the graph Laplacian, but also incorporates the label information as the fitting constraints to learn.
- 25
- Partitioning networks into communities by message passing23
The successful validation of the proposed scheme shows that this approach can effectively address the problem of resolution limit and find communities in both directed and undirected networks within a unified framework, including revealing multiple levels of robust community partitions.
- Symmetric Graph Regularized Constraint Propagation22
It is clearly shown that pairwise constraint propagation is actually equivalent to solving a Lyapunov matrix equation, which is widely used in Control Theory as a standard continuous-time equation.
- 22
- Structural and Statistical Texture Knowledge Distillation and Learning for Segmentation21
A novel Structural and Statistical Texture Knowledge Distillation (SSTKD) framework for semantic segmentation and a Co-occurrence TIEM (C-TIEM) and generic segmentation frameworks to enable existing segmentation networks to harvest the structural and statistical texture information more effectively.
- Discrete Latent Perspective Learning for Segmentation and Detection21
A novel framework, Discrete Latent Perspective Learning (DLPL), is proposed for latent multi-perspective fusion learning using conventional single-view images to achieve consistent semantic interpretation in machine learning and computer vision.
- Exponential synchronization of weighted general delay coupled and non-delay coupled dynamical networks21
This work investigates the exponential synchronization of weighted general delay coupled and non-delay coupled complex dynamical networks with different topological structures based on the Lyapunov stability theory, and proves theoretically the effectiveness and feasibility of the synchronization scheme.
- Contrast, Stylize and Adapt: Unsupervised Contrastive Learning Framework for Domain Adaptive Semantic Segmentation20
This work presents CONtrastive FEaTure and pIxel alignment (CON-FETI) for bridging the domain gap at both the pixel and feature levels using a unique contrastive formulation and demonstrates that this method outperforms existing state-of-the-art methods using DeepLabV2.
- Optimal Watermark Detection Based on Support Vector Machines20
A novel optimal watermark detection scheme based on support vector machine and error correcting codes is proposed to extract the watermark bits from a possibly corrupted marked image with a lower error probability.
- Changenet: Multi-Temporal Asymmetric Change Detection Dataset19
The ChangeNet dataset consists of 31,000 multi-temporal images pairs, a wide range of complex scenes from 100 cities, and 6 pixel-level annotated categories, which is far superior to all the existing change detection datasets including LEVIR-CD, WHU Building CD, etc..
- 19
- 19
- Novel robust image watermarking using difference correlation detector19
A robust digital image watermarking scheme based on subsampling and difference correlation that has a better performance than a common coefficient correlation detection and does not need the original image.
- Global exponential convergence of multitime-scale neural networks19
A theoretic condition for global exponential convergence of the solutions of the network is provided, which is proved weaker than some existing results in the literature.
- 18
- Adaptive Overlapping Community Detection with Bayesian NonNegative Matrix Factorization18
Non-negative Matrix Factorization for Community Detection using Bayesian explanation in factorizing process can not only catch most appropriate count of communities in a large network with Shrinkage method, but also verify good threshold how a node should be assigned to a community in fuzzy situation.
- DirectMHP: Direct 2D Multi-Person Head Pose Estimation with Full-range Angles17
This paper designs a novel end-to-end trainable one-stage network architecture by joint regressing locations and orientations of multi-head to address the MPHPE problem, and regards pose as an auxiliary attribute of the head, and append it after the traditional object prediction.
- Body-Part Joint Detection and Association via Extended Object Representation17
A novel extended object representation is proposed that integrates the center location offsets of body or its parts, and a dense single-stage anchor-based Body-Part Joint Detector (BPJDet) is constructed that can be seamlessly generalized to jointly detect any body part.
- Completely Heterogeneous Federated Learning17
This work proposes the challenging"completely heterogeneous"scenario in FL, which refers to that each client will not expose any private information including feature space, model architecture, and label distribution, and devise an FL framework based on parameter decoupling and data-free knowledge distillation to solve the problem.
- End to end multi-scale convolutional neural network for crowd counting17
The proposed model extract multi-scale deep CNN features from the input image and regress the crwod count directly, without any post-processing, which could handle muti-scale targets well in various crowd scene.
- Non-negative Matrix Factorization with Pairwise Constraints and Graph Laplacian17
This paper proposes a novel pairwise constrained non-negative matrix factorization with graph Laplacian method, which not only utilizes the local structure of the data by graph LaPLacian, but also incorporates pairwise constraints generated among all labeled data into NMF framework.
- Revealing digital fakery using multiresolution decomposition and higher order statistics17
Experiments show that the proposed detection scheme is effective, demonstrating that the suggested statistical features can model the differences between natural images and fake images.
- 17
- RAFT-MSF: Self-Supervised Monocular Scene Flow Using Recurrent Optimizer16
This paper designs a new decoder to iteratively update 3D motion fields and disparity maps simultaneously based on RAFT, a state-of-the-art optical flow model, and proposes an enhanced upsampling layer and a disparity initialization technique that substantially improves accuracy.
- Image classification based on improved VLAD16
Experimental results on UIUC Sports Event, Corel 10 and 15 Scenes datasets show that the proposed methods outperform some state-of-the-art coding schemes in terms of the classification accuracy and computation speed.
- 16
- Extracting weights from edge directions to find communities in directed networks16
This work addresses the problem of community detection in directed networks by extracting information about edge directions using a PageRank random walk and translating such information into edge weights, which results in a new weighted directed network in which edge directions can be safely ignored.
- YOLO-Med : Multi-Task Interaction Network for Biomedical Images15
This study proposes an efficient end-to-end multi-task network capable of concurrently performing object detection and semantic segmentation called YOLO-Med, which employs a backbone and a neck for multi-scale feature extraction and a cross-scale task-interaction module is employed in order to facilitate information fusion between various tasks.
- 15
- Linear discriminant analysis with spectral regularization15
This paper proposes trace norm regularized LDA that not only tackles the problem of small sample size but also uncover the underlying structures between target classes through the appealing property of trace norm.
- 15
- BECAME: BayEsian Continual Learning with Adaptive Model MErging14
The potential of model merging to enhance the stability-plasticity trade-off is explored, providing theoretical insights that underscore its benefits and introducing a two-stage framework named BECAME, which synergizes the expertise of gradient projection and adaptive merging.
- DAG: Deep Adaptive and Generative K -Free Community Detection on Attributed Graphs14
This paper proposes a novel Deep Adaptive and Generative model~(DAG) for community detection without specifying the prior K, and designs a new metric, EDGE, to evaluate community detection methods even when the labels are not feasible.
- Enhancing the Accuracy of Lymph-Node-Metastasis Prediction in Gynecologic Malignancies Using Multimodal Federated Learning: Integrating CT, MRI, and PET/CT14
It is suggested that federated learning, especially when combined with MRI assessment data, can revolutionize lymph-node-metastasis detection in gynecological malignancies and pave the way for more precise patient care, potentially transforming the current diagnostic paradigm and resulting in improved patient outcomes.
- Deep Learning Assisted Diagnosis of Musculoskeletal Tumors Based on Contrast‐Enhanced Magnetic Resonance Imaging14
Misdiagnosis of malignant musculoskeletal tumors may lead to the delay of intervention, resulting in amputation or death, and it is necessary to select patients suitable for amputation.
- 14
- UNIDEAL: Curriculum Knowledge Distillation Federated Learning13
The proposed method introduces Adjustable Teacher-Student Mutual Evaluation Curriculum Learning, which significantly enhances the effectiveness of knowledge distillation in FL settings, and achieves superior performance in terms of both model accuracy and communication efficiency.
- Stuart: Individualized Classroom Observation of Students with Automatic Behavior Recognition And Tracking13
StuArt, a novel automatic system designed for the individualized classroom observation, which empowers instructors to concern the learning status of each student, is presented.
- 13
- BPJDet: Extended Object Representation for Generic Body-Part Joint Detection12
A novel extended object representation integrating center-offsets of body parts is proposed, and an end-to-end generic Body-Part Joint Detector (BPJDet) is constructed, which can be generalized to detect body-part or body-parts of either human or quadruped animals.
- Library book recommendation with CNN-FM deep learning approach12
A novel embedding-driven model called InFo is proposed, which refers to users' intrinsic interests and academic preferences to provide personalized library book recommendations and outperforms other state-of-the-art methods in terms of two metrics called Recall and NDCG.
- Neighbor similarity and soft-label adaptation for unsupervised cross-dataset person re-identification12
A Neighbor Similarity and Soft-label Adaptation algorithm to transfer the supervised information from source domain to a new unlabeled target dataset and significantly outperforms the state-of-the-art unsupervised and domain adaptation re-identification methods.
- Stylized Adversarial AutoEncoder for Image Generation12
An autoencoder-based generative adversarial network (GAN) for automatic image generation, which is called "stylized adversarial autoenCoder", which achieves superior results for image generation as well as remarkably improves the corresponding supervised recognition task.
- LEARNING OVERLAPPING COMMUNITIES IN COMPLEX NETWORKS VIA NON-NEGATIVE MATRIX FACTORIZATION12
This approach effectively estimates the number of communities and is more interpretable and more accurate in identifying overlapping communities in undirected networks than previous approaches.
- Robust object tracking with occlusion handle12
A novel incremental subspace (robust PCA)-based object tracking algorithm to deal with the occlusion problem and the introduction of incremental PCA facilitates the subspace updating process and possesses several benefits compared with traditional R-SVD-based updating methods.
- Absolute exponential stability analysis of delayed neural networks12
This Letter investigates the absolute exponential stability of a class of delayed neural networks and generalizes the additive diagonal stability condition of the interconnection weight matrix to the case of neural networks with delays.
- Comments on "A generalized LMI-based approach to the global asymptotic stability of delayed cellular neural networks"12
It is shown that the linear matrix inequality (LMI)-based criterion obtained for the global exponential stability of the delayed neural networks can be simplified to a simpler but equivalent form and, thus, it is not necessary to have such complex form of condition in the above paper.
- Federated Multi-Task Learning on Non-IID Data Silos: An Experimental Study11
This paper introduces a novel framework, FMTL-Bench, for systematic evaluation of the FMTL paradigm, and comprises seven sets of comparative experiments, encapsulating a wide array of non-independent and identically distributed data partitioning scenarios.
- Phytochemical interventions for post-traumatic stress disorder: A cluster co-occurrence network analysis using CiteSpace11
Research on phytochemicals and PTSD is unevenly distributed across countries/regions, disciplines, and journals, leading to the exploration of botanical active ingredients and molecular mechanisms.
- Degradation-Guided Meta-Restoration Network for Blind Super-Resolution11
A Degradation-guided Meta-restoration network for blind Super-Resolution (DMSR) that facilitates image restoration for real cases and outperforms SOTA by a large margin on three widely-used benchmarks is proposed.
- FH-GAN: Face Hallucination and Recognition Using Generative Adversarial Network11
A Face Hallucination Generative Adversarial Network (FH-GAN) is proposed which improves the quality of low resolution face images and accurately recognize those low quality images and a new face hallucination network, namely Dense Sparse Network (DSNet), which improves upon the state-of-art in face hallucinationsination.
- Local similarity learning for pairwise constraint propagation11
This paper makes attempt to exploit the available pairwise constraints to learn a new set of similarities, which are consistent with the supervisory information in the Pairwise constraints, before propagating these initial constraints.
- An integrated Gaussian mixture model to estimate vigilance level based on EEG recordings11
An integrated approach to estimate vigilance level is proposed, which incorporates an automatically artifact removing preprocess, a novel vigilance labeling method and finally a Gaussian Mixed Model (GMM) to discover the underlying pattern of EEG signals.
- 11
- Group Sparse Non-negative Matrix Factorization for Multi-Manifold Learning11
This paper proposes a novel algorithm to learn multiple linear manifolds for face recognition, called Group Sparse Non-negative Matrix Factorization (GSNMF), which achieves accurate recognition on face images with varying illuminations and expressions.
- IDENTIFICATION OF COMMUNITY STRUCTURE IN COMPLEX NETWORKS USING AFFINITY PROPAGATION CLUSTERING METHOD11
Based on a new method for calculating similarity between pairs of vertices and a transforming method for a given similarity from likelihood to log-domain, an affinity propagation clustering method is applied to identify communities in complex networks.
- Color Image Watermarking Based on Neural Networks11
Experimental results show good performance of the proposed scheme resisting common signal processing and geometric attacks.
- Cooperative Self-Training for Multi-Target Adaptive Semantic Segmentation10
This work proposes a self-training strategy that employs pseudo-labels to induce cooperation among multiple domain-specific classifiers and employs feature stylization as an efficient way to generate image views that forms an integral part of self- training.
- 10
- Joint Multi-Person Body Detection and Orientation Estimation via One Unified Embedding10
This work proposes a single-stage end-to-end trainable framework for tackling the HBOE problem with multi-persons by integrating the prediction of bounding boxes and direction angles in one embedding into the multi-scale anchor channel predictions of persons for concurrently benefiting from engaged intermediate features.
- Semi-deterministic and Contrastive Variational Graph Autoencoder for Recommendation10
This paper proposes a novel Semi-deterministic and Contrastive Variational Graph autoencoder (SCVG) for item recommendation, and empirically shows that the contrastive regularization makes learned user/item latent representation more personalized and helps to smooth the training process.
- 10
- A Novel Watermark Algorithm for Integrity Protection of XML Documents10
This paper proposes a novel solution to the integrity protection of XML documents by generating watermarks through applying Principal Component Analysis on a matrix constructed from an XML document and a key and embedding them into the document by altering the case of letters in the XML tags.
- GA-Driven LDA in KPCA Space for Facial Expression Recognition10
A rather simple but effective preprocessing method using Mahalanobis distance rather than Euclidean distance as the metric of the nearest neighbor classifier and experiments on the JAFFE database show promising results.
- Generating Negative Samples for Multi-Modal Recommendation9
This paper proposes NegGen, a novel framework that utilizes multi-modal large language models (MLLMs) to generate balanced and contrastive negative samples and employs a causal learning module to disentangle the effect of intervened key features and irrelevant item attributes, enabling fine-grained learning of user preferences.
- Position-Aware Subgraph Neural Networks with Data-Efficient Learning9
A novel node position encoding method that is anchor-free is proposed, and a new generative subgraph augmentation method based on a diffused variational subgraph autoencoder is designed, and exploratory and exploitable views for subgraph contrastive learning are proposed.
- Trimap-guided feature mining and fusion network for natural image matting9
A trimap-guided feature mining and fusion network consisting of the TMP module and GLF modules, which focuses effective feature mining on interesting objects under the guidance of trimap without extra parameters, and a common interesting object matting dataset to advance high-quality image matting.
- 9
- 9
- Action Recognition Using Hybrid Feature Descriptor and VLAD Video Encoding9
A hybrid feature descriptor is proposed, which combines both static descriptor and motional descriptor to cover more action information inside video clips, and facilitated with VLAD encoding method, outperforms traditional descriptors by a large margin.
- 9
- Absolute exponential stability of a class of recurrent neural networks with multiple and variable delays9
Using the Holder's inequality and the Young's inequality to estimate the derivatives of the Lyapunov functionals, some new conditions for absolute exponential stability (AEST) of a class of recurrent neural networks with multiple and variable delays are derived.
- Towards Dynamic and Scalable Active Learning with Neural Architecture Adaption for Object Detection8
A novel active learning strategy with dynamic neural architecture adaption for object detection that avoids the retraining of the detector after changing the network architecture in each round, and introduces Dirichlet calibration to correct the classifier for obtaining credible prediction.
- 8
- Locality Preserving Discriminative Hashing8
A novel semi-supervised hashing method called Locality Preserving Discriminative Hashing which combines two classical dimensionality reduction approaches, Linear Discriminant Analysis (LDA) and LocalityPreserving Projection (LPP).
- Multi-label Linear Discriminant Analysis with Locality Consistency8
A new method is presented called Multi-label Linear Discriminant Analysis with Locality Consistency (MLDA-LC) which incorporates local structure into the framework of MLDA and employs a graph regularized term to preserve the local structure for multi-label data.
- Head Pose Estimation Based on Manifold Embedding and Distance Metric Learning8
This paper uses a learned Mahalanobis distance to seek optimal subjects with similar manifold to construct the embedding and proposes a new smooth and discriminative embedding method supervised by both pose and identity information.
- Smooth Multi-Manifold Embedding for Robust Identity-Independent Head Pose Estimation8
This paper considers the pose data space as multiple manifolds in which each manifold characterizes the underlying subspace of subjects with similar appearance and proposes a novel embedding criterion to learn each manifold from the exemplar-centered local structure of subjects.
- 8
- Robust Incremental Subspace Learning for Object Tracking8
The proposed incremental subspace based object tracking algorithm is efficient and effective to cope with common object tracking tasks, especially with strong robustness due to the introduction of Robust PCA.
- Evolutionary Discriminant Feature Extraction with Application to Face Recognition8
A new evolutionary approach to extracting discriminant features with low space complexity and high search efficiency is proposed, which provides an effective way to investigate the discriminative power of different feature subspaces.
- A fast evolutionary pursuit algorithm based on linearly combining vectors8
This paper proposes a fast Evolutionary pursuit algorithm that searches for optimal discriminant vectors through linearly combining standard bases and verifies its effectiveness and enhanced efficiency.
- Towards Personalized Federated Multi-Scenario Multi-Task Recommendation7
This paper proposes PF-MSMTrec, a novel framework for personalized federated multi-scenario multi-task recommendation using the Multi-gate Mixture-of-Experts (MMoE) structure, and introduces a bottom-up joint learning mechanism to address the unique challenges of multiple optimization conflicts.
- 7
- 7
- ULN: An efficient face recognition method for person wearing a mask7
This work generates numerous facial images with masks based on public face datasets and proposes a novel network architecture called Upper-Lower Network (ULN) to recognize the faces with masks efficiently and outperforms other state of the art face recognition methods.
- Improve Image Captioning by Self-attention7
Extensive experiments on MSCOCO dataset show that the self-attention can empower the captioning model to achieve competitive performance with the state-of-the-art methods.
- 7
- Generalized residual vector quantization for large scale data7
This paper proposes a novel vector quantization framework that iteratively minimizes quantization error and substantially outperforms existing methods in terms of quantization accuracy and computation efficiency.
- Image classification based on nearest neighbor basis vectors7
This paper exploits the widely used Spatial Pyramid Matching model (SPM) in the paper and proposes an image classification method based on nearest neighbor basis vectors of the dictionary, which outperforms some state-of-the-art methods.
- 7
- 7
- Automatic Feature Extraction and Analysis on Breast Ultrasound Images7
The experimental results showed that the individual morphologic feature measure had the strong ability to distinguish the malignant and benign breast lesions, especially the StdLAI, and SnrLAI could reach 0.92 with the specificity, while the discrimination performance of individual feature measure on margin, texture and acoustic shadowing was relatively low on the database.
- From Anchors to Answers: A Novel Node Tokenizer for Integrating Graph Structure into Large Language Models6
NT-LLM is presented, a novel framework with an anchor-based positional encoding scheme for graph representation that enhances LLMs' ability to understand and reason with graph-structured information, offering an efficient solution for graph-based applications of language models.
- Enhancing Cross-Domain Detection: Adaptive Class-Aware Contrastive Transformer6
A novel class-aware cross domain detection transformer based on the adversarial learning and mean-teacher framework is proposed, which outperforms the state-of-the-art transformer based methods in improving performance and alleviating class imbalance issues.
- Task-Interaction-Free Multi-Task Learning with Efficient Hierarchical Feature Representation6
This study introduces a Task-Interaction-Free Network (TIF) for multi-task learning, which diverges from explicitly designed task interaction mechanisms and proposes task and scale-specific decoders that efficiently decode the enhanced features shared across tasks without necessitating task-interaction modules.
- Panet: Perspective-Aware Network with Dynamic Receptive Fields and Self-Distilling Supervision for Crowd Counting6
This paper proposes a novel perspective-aware approach called PANet, which outperforms the state-of-the-art methods by a large margin and proposes the dynamic receptive fields (DRF) framework, which is able to adjust the receptive field by the dilated convolution parameters according to the input image.
- Arbitrary shape scene text detector with accurate text instance generation based on instance-relevant contexts6
An arbitrary shape text detector that learns and combines instance-relevant contexts to generate accurate text instances of different patterns in scene images and effectively improves the text detector’s detection accuracy and robustness ability is proposed.
- 6
- AT-Text: Assembling Text Components for Efficient Dense Scene Text Detection6
A novel Assembling Text Components (AT-text) that accurately detects dense text in scene images and achieves satisfactory results and a considerable balance between precision and recall without a large margin in ICDAR2013 and MSRA-TD 500 public benchmark datasets.
- Community detection in scientific collaborative network with bayesian matrix learning6
This work considers that author u and v pay same contribution for one paper, and whole collaborative network is undirected, and community detection with nonnegative matrix factorization (NMF) can capture the underlying structure of network in the low dimensional data space with its community-based representation, and achieve good performance in many scientific collaborative networks.
- Novel multi-tap analog self-interference cancellation architecture with shared phase-shifter for full-duplex communications6
In the proposed architecture, all taps share one phase shifter to emulate the dominated phase offset of the SI channel, which reduces the complexity of the implementation of the multi-tap analog SIC structure and avoids the SIC performance degradation.
- Graph regularized tensor sparse coding for image representation6
A novel graph regularized tensor sparse coding for image representation that preserves the local proximity of elementary structures in the image by adopting the newly proposed tubal-tensor representation and considers the intrinsic geometric properties by imposing graph regularization.
- 6
- Online Recommender System Based on Social Network Regularization6
An online collaborative filtering algorithm which can better use the social information and prior knowledge is proposed and it is proposed that this problem of cold start and users with few ratings is solved.
- AUTOMATIC DEFECT CLASSIFICATION OF TFT-LCD PANELS WITH SHAPE, HISTOGRAM AND COLOR FEATURES6
A novel LBP-E feature considering intensity equality proposed in this paper is compared to other original rotation invariant LBP (Local Binary Pattern) features and shows that the method can generate a better result with a relatively low dimension number.
- Classification of Alzheimer’s Disease Based on Cortical Thickness Using AdaBoost and Combination Feature Selection Method6
This research designed and applied a new supervised learning algorithm for classification of Alzheimer’s disease (AD) using MRI cortical surface-based analysis, and combined all the classifiers to form a more powerful classifier using AdaBoost.
- 6
- A LOCAL INFORMATION-BASED ROUTING STRATEGY ON THE SCALE-FREE NETWORK6
This paper proposes a new routing strategy which is only based on local information of network topology which integrated the delivering capability and packets queue length of nodes for enhancing the efficiency of traffic on scale-free networks.
- Detecting fake images using watermarks and support vector machines6
A detection scheme, which, firstly, uses watermarks to locate the alteration, and then, uses support vector machine (SVM) as a classifier to make a binary decision on whether an image is fake or real is proposed.
- A direct evolutionary feature extraction algorithm for classifying high dimensional data6
The proposed direct evolutionary feature extraction algorithm for classifying high-dimensional data has an excellent performance in classification while reducing the space complexity by an order of magnitude and is compared with some representative feature extraction algorithms in face recognition.
- CHAOTIC PHASE SHIFT KEYING IN DELAYED CHAOTIC ANTICONTROL SYSTEMS6
A new chaotic phase shift keying (CPSK) scheme is proposed for secure communications based on the delayed feedback chaotic anticontrol systems and can exhibit rich chaotic behavior with the choice of appropriate parameters.
- 6
- 5
- 5
- A Robust and Effective Text Detector Supervised by Contrastive Learning5
An effective and robust text detection network that combines a state-of-the-art contrastive learning method SimCLR and then calculates the similarity of the extracted corresponding feature pairs to significantly improve the performance of the detector in difficult conditions is proposed.
- 5
- 5
- 5
- On Multi-modal Fusion Learning in constraint propagation5
The fusion learning and constraint propagation into one unified problem and solve it by a bound-constrained quadratic optimization is presented, which can reach a multi-modal fusion based on the observed constraint information and the propagation process.
- Convolutional Gated Recurrent Units Fusion for Video Action Recognition5
This work introduces a Convolutional Gated Recurrent Units (ConvGRU) fusion method to model long-term dependency inside actions and proposes an end-to-end architecture according to this fusion method which is comparable to the state-of-the-art.
- 5
- 5
- Constrained Non-negative Matrix Factorization with Graph Laplacian5
This paper presents Constrained Non-negative Matrix Factorization with Graph Laplacian (CNMF-GL), which not only employs the geometrical information, but also properly uses the label information to enhance NMF.
- 5
- Modalities consensus for multi-modal constraint propagation5
This paper first combines multiple single-modal constraint propagation problems together, and then explicitly introduces a new modalities consensus regularizer to force the propagation results on different modalities to be consistent with each other.
- An Integrated Hierarchical Gaussian Mixture Model to Estimate Vigilance Level Based on EEG Recordings5
A new integrated approach to predict human vigilance level is proposed, which incorporate an automatically artifact removing pre-process, a novel vigilance quantification method and finally a hierarchical Gaussian Mixed Model(hGMM) to discover the underlying pattern of EEG signals.
- Incremental visual objects clustering with the growing vocabulary tree5
A novel approach to cluster visual objects in an incremental manner by adapting to the visual domain of the incremental pLSA model previously used for text analysis by expanding the visual vocabulary to include the new visual words, and adjusting the objects clustering model to absorb these new words.
- 5
- 5
- Fragile Watermarking Schemes for Tamperproof Web Pages5
Two novel fragile watermarking schemes for tamperproof web pages are proposed, in which a message authentication method and a sparse coding method are used to generate the watermarks for a web page, respectively.
- 5
- 5
- 5
- GLOBAL SYNCHRONIZATION IN AN ARRAY OF LINEARLY COUPLED DELAYED NEURAL NETWORKS WITH AN ARBITRARY COUPLING MATRIX5
This paper considers an array of linearly coupled identical delayed neural networks with an arbitrary coupling matrix without assuming it to be symmetric, irreducible and diffusive, and derives a new sufficient condition for global synchronization.
- Convergence Analysis of Genetic Regulatory Networks Based on Nonlinear Measures5
A nonlinear 2-measure concept is proposed, together with the nonlinear 1-measures proposed earlier by other researchers, to analyze the global convergence of genetic regulatory networks and provides insight into the dynamical behavior of gene networks.
- Subsampling-Based Robust Watermarking Using Neural Network Detector5
Experimental results show good performance of the proposed scheme on resisting common signal processing attacks, and the neural network detector can nearly exactly extract the payload watermark.
- T‐Skeleton: Accurate scene text detection via instance‐aware skeleton embedding4
This paper proposes an efficient text detector, namely T‐Skeleton, to obtain more reliable segmentation detections and introduces the text instance skeleton highlighting the semantically significant structure to explicitly capture the long‐range dependencies of text instances.
- 4
- 4
- 4
- 4
- 4
- Nutritional supplements formulated to prevent cognitive impairment in animals4
The results demonstrated that certain functional food factors and two designed formulations significantly prevent cognitive impairment of heat-stressed animals.
- 4
- Multi-modal constraint propagation via compatible conditional distribution reconstruction4
This paper first identifies a necessary and sufficient condition for compatible conditional distributions under a specific assumption and address the problem of Compatible Conditional Distributions Reconstruction (CCDR), and proposes a multi-modal constraint propagation method dubbed Instance Level Multi-Modal Constraint Propagation (ILMCP).
- 4
- Local Linear Spectral Hashing4
This paper proposes a \(local\ linear\ spectral\ hashing\) framework that minimizes the average Hamming distance with a new local neighbor matrix, which can guarantee the mapping not only from neighbor images to neighbor codes, but also from non-neighbor images to non-NEighbor codes.
- 4
- 4
- VIGILANCE ANALYSIS BASED ON EEG SIGNALS: SEEKING FOR SUITABLE FEATURES4
This paper uses EEG recordings to discriminate three vigilance states of a person, namely alert, drowsy, and sleep, while driving a car in a simulation environment and investigates and compares three different kinds of features which are based on Continuous Wavelet Transform, Discrete Wavelet transform, and Fractal Dimension, respectively.
- Analysis of R-D Location Precision in Spaceborne SAR Image4
Analysis and computation of SAR location error are carried out and simulation results indicate that the distance measurement error, terrain variation error, and the satellitic situation and speed error are the main factors influencing the location precision, but the influence caused by Doppler frequency measurement error can be ignored.
- TRAFFIC CONGESTION ANALYSIS IN COMPLEX NETWORKS BASED ON VARIOUS ROUTING STRATEGIES4
This paper investigates numerically the problem of traffic congestion in complex networks through the use of various routing strategies, and evaluates the efficiency of different routing strategies on different network topologies and shows how the network structures and routing strategies influence the traffic congestion status.
- Robust Image Watermarking Using RBF Neural Network4
A digital image watermarking scheme is developed using neural network to embedded watermark into DCT domain of each subimage blocks obtained by subsampling, which achieves adaptively watermark embedding and stronger robustness.
- Color Image Watermarking Based on Amplitude Modulation4
A fixed binary sequence is added at the head of the payload watermark as a part of the embedded watermark sequence which can make an adaptive threshold in watermark extraction.
- Edge-side NPU inference optimization: Adaptation research of multimodal large models on qualcomm platforms3
In-depth adaptation and optimization of multimodal large models on the Neural Network Processing Unit (NPU) based on the Qualcomm platform is conducted, significantly improving the model’s response speed and decoding efficiency.
- Task Indicating Transformer for Task-Conditional Dense Predictions3
This paper introduces a novel task-conditional framework called Task Indicating Transformer (TIT), which incorporates a Task Indicating Matrix through matrix decomposition, thereby enhancing long-range dependency modeling and parameter-efficient feature adaptation by capturing intra- and inter-task features.
- Deep convolutional neural network model ResNeSt for discrimination of papillary thyroid carcinomas and benign nodules in thyroid nodules diagnosed as atypia of undetermined significance3
The DCNN model ResNeSt showed great potential in discriminating thyroid nodules diagnosed as AUS, and among those nodules, malignant nodules showed larger and more deeply stained nuclei than benign nodules.
- Beyond Binary Preference: Leveraging Bayesian Approaches for Joint Optimization of Ranking and Calibration3
This paper proposes an effective data augmentation strategy, named Beyond Binary Preference (BBP), which significantly outperforms state-of-the-art methods in both ranking and calibration capabilities, showcasing its effectiveness in addressing the limitations of existing methods.
- 3
- An Intuitive and Unconstrained 2D Cube Representation for Simultaneous Head Detection and Pose Estimation3
A novel single-stage keypoint-based method via an intuitive and unconstrained 2D cube representation for joint head detection and pose estimation that can effortlessly obtain the closed-form solution of Euler angles from predicted 2D head cube instead of applying the error-prone PnP algorithm.
- 3
- 3
- 3
- 3
- 3
- Multilevel Collaborative Attention Network for Person Search3
This work proposes a novel Multilevel Collaborative Attention Network (MCAN) to fulfill person search task efficiently, and achieves superiority over current state-of-the-art methods.
- Image Recognition with Deep Learning for Library Book Identification3
A new image recognition method to identify books in libraries based on barcode decoding together with deep learning optical character recognition (OCR) and its application in library book identification system is described.
- 3
- 3
- LSOD3
It is shown that the proposed descriptor is not only invariant to geometric and photometric transformations (such as viewpoint change, intensity change, noise, image blur and JPEG compression), but also highly efficient.
- 3
- 3
- 3
- A Robust Blind Image Watermarking Scheme Based on Template in Lab Color Space3
This scheme is a blind watermarking scheme, extraction can be proposed without the presence of original image, and experiments show that the scheme is resilient to typical image processing like cropping, scaling, and translation.
- 3
- 3
- SVR-Based Oblivious Watermarking Scheme3
Experimental results show that the proposed SVR based oblivious watermarking scheme for color image have outperforming performance over Kutter's method against several different attacks including noise addition, shearing, luminance & contrast enhancement, distortion, etc.
- How Does Topology Bias Distort Message Passing in Graph Recommender? A Dirichlet Energy Perspective2
Test-time Simplicial Propagation (TSP), which extends message passing to higher-order simplicial complexes and mitigates harmful topology bias and substantially improves the representation and recommendation of long-tail items during inference.
- 2
- Few-Shot Implicit Function Generation via Equivariance2
Experiments on 2D image and 3D shape INR datasets demonstrate that the EquiGen approach effectively generates diverse INR weights while preserving their functional properties in few-shot scenarios.
- 2
- 2
- Combining Past, Present and Future: A Self-Supervised Approach for Class Incremental Learning2
A self-supervised CIL framework CPPF, meaning Combining Past, Present and Future, is proposed, which boosts the performance of self- supervised class incremental learning.
- Radiologists with assistance of deep learning can achieve overall accuracy of benign–malignant differentiation of musculoskeletal tumors comparable with that of pre-surgical biopsies in the literature2
Use of DL algorithms on musculoskeletal dynamic contrast-enhanced MRI improved the benign–malignant differentiation accuracy of radiologists to a level comparable to that of pre-surgical biopsies.
- NoMorelization: Building Normalizer-Free Models from a Sample's Perspective2
NoMorelization is a simple and effective alternative to normalization which is composed of two trainable scalars and a zero-centered noise injector and shows the best speed-accuracy trade-off.
- Generative Adversarial Networks with Bi-directional Normalization for Semantic Image Synthesis2
Bi-directional Normalization (BDN) is proposed in the authors' generative adversarial networks to solve problems of semantic label information and real scene image feature representation to be effectively utilized by a bi- directional way for generating high quality images.
- 2
- Small and accurate heatmap-based face alignment via distillation strategy and cascaded architecture2
A face alignment model with 3.64M parameters is obtained, which is about 6x smaller than the cumbersome model, and outperforms the state-of-the-art algorithms on both AFLW2000-3D and 300W-LP.
- Denstity Level Aware Network for Crowd Counting2
A novel method called Density Level Aware Network (DLA-Net) to improve the density map estimation in varying density scenes by dividing the input into multiple regions according to their density levels and handling the regions independently.
- A Multimodal Lossless Coding Method for Skeletons in Videos2
This paper proposes a multi-modal skeleton coding tool containing three different coding schemes, namely, spatial differential-Coding scheme, motion-vector-based differential-coding scheme and inter prediction scheme, thus utilizing both spatial and temporal redundancy to losslessly compress skeleton data.
- 2
- Impact of the Amount of RF Self-interference Cancellation on Digital Self-interference Cancellation in Full Duplex Communications2
The experimental results in this paper show that the effectiveness of digital self-interference cancellation decreases with the increase in the amount of RF self- Interference cancellation and digital cancellation is more effective when applied selectively based on measured values.
- Stereo Matching Using Conditional Adversarial Networks2
The first application of conditional adversarial networks to stereo matching task is presented, and high-level contextual features are incorporated into loss function in order to help the generator to correct predicted disparity maps.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Sparse Deep Belief Net for Handwritten Digits Classification2
Another version of Sparse Deep Belief Net is proposed which applies the differentiable sparse coding method to train the first level of the deep network, and then train the higher layers with RBM, which leads to state-of-the-art performance on the classification of handwritten digits.
- 2
- 2
- Satellite Interference Locating Method based on Support Vector Regression2
A new satellite interference locating method based on support vector regression and the satellite multi-beam antennas model is presented that can achieve higher accurate estimates for direction of arrival (DOA) while avoiding the all-direction peak value searching technique used in other traditional DOA estimation methods.
- Parsimonious Feature Extraction Based on Genetic Algorithms and Support Vector Machines2
The proposed parsimonious feature extraction algorithm searches for the optimal features using a genetic algorithm and evaluates the features referring to Support Vector Machines to achieve the same or even better classification performance.
- Cervi-LLM for real time colposcopy lesion detection and interpretable diagnosis1
Cervi-LLM effectively overcomes conventional colposcopy limitations, offering an accurate, real-time intelligent tool for cervical lesion screening and biopsy guidance.
- 1
- MORE: Multi-Organ Medical Image REconstruction Dataset1
The Multi-Organ medical image REconstruction (MORE) dataset is introduced, comprising CT scans across 9 diverse anatomies with 15 lesion types, enabling robust training of deep learning models on extensive, heterogeneous data and facilitating rigorous evaluation of model generalization for CT reconstruction.
- 1
- Evidence map of nutrient interventions for major depression: a systematic review protocol1
This study aims to systematically evaluate evidence for various nutritional interventions in depression and create an evidence gap map (EGM) to inform individualized nutritional strategies, improve clinical guidelines, and potentially influence public health policies.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- EDEN: A Plug-in Equivariant Distance Encoding to Beyond the 1-WL Test1
This work theoretically proves that EDEN is permutation-equivariant for all level graph representation learning, and empirically illustrates that EDen's expressive power can reach up to the 3-WL test.
- 1
- 1
- 1
- 1
- 1
- Missing Frame Detection of Surveillance Videos Based on Deep Learning in Forensic Science1
A method to detect missing frames by the movement of objects in video based on object detection neural network that effectively reduces calculation errors of vehicle speed and promotes the authenticity of forensic investigation is proposed.
- Temporal Continuity Based Unsupervised Learning for Person Re-identification1
Temporal Continuity based Unsupervised Learning (TCUL) simultaneously does center based clustering of unlabeled dataset and fine-tunes a convolutional neural network pre-trained on irrelevant labeled dataset to enhance discriminative capability of the CNN for the target dataset.
- Community Detection in Knowledge Graph Network with Matrix Factorization Learning1
This paper investigates some main methods of current network analysis and community detection tasks related with knowledge graph and semantic network, and achieves good performance for community detection in various semantic network data.
- Recurrent Spatiotemporal Feature Learning for Action Recognition1
The proposed architecture is end-to-end trainable, and of significant flexibility to be adapted in any CNN-based structure, and produces the state-of-the-art performance on two standard benchmark for action recognition over RNN-based approaches.
- 1
- Orthogonal optimal reverse prediction for semi-supervised learning1
This paper proposes an orthogonal optimal reverse prediction (OORP), together with its kernelized and Laplacian regularized extensions, which can greatly enhance the classification performance, compared to the original ORP.
- 1
- 1
- 1
- Vigilance Estimation Based on Statistic Learning with One ICA Component of EEG Signal1
Inspired by the HMM(Hiden Markov Model), a probability method based on the (PSD) Power Spectral Density distribution of the energy changes to estimate the vigilance level of humans using only one ICA(Independent Component Analysis) component of EEG signal is proposed.
- A novel web application tamper proof method1
This paper compares the new proposed system with intrusion detection system, and analyzes why this system can prevent more kinds of malevolence attack.
- A novel blinding digital watermark algorithm based on lab color space1
This paper presents an innovative blinding color image watermark algorithm based on Lab color space, which first marks the watermark region size and position through embedding some regular blocks called anchor points in image spatial domain, and then embeds the watermarks into the image.
- Large Scale Visual Classification via Learned Dictionaries and Sparse Representation1
The experimental results demonstrate that the proposed K-SVD algorithm outperforms k-nearest-neighbor in two aspects: 1) the discriminative capability for large scale visual classification task, and 2) the average running time of classifying one image.
- 1
- EFFICIENT ROUTING STRATEGY ON SCALE-FREE NETWORKS1
A new routing strategy with two tunable parameters based on local information of network, which includes the static information, node degree, and dynamic information, buffer queue length, to guide the selection of the next transmission node.
- 1
- 1
- 1
- 1
- –
- Recovering Hidden Reward in Diffusion-Based Policies–
EnergyFlow is introduced, a framework that unifies generative action modeling with inverse reinforcement learning by parameterizing a scalar energy function whose gradient is the denoising field, and it is proved that constraining the learned field to be conservative reduces hypothesis complexity and tightens out-of-distribution generalization bounds.
- HACMatch: Semi-supervised rotation regression with hardness-aware curriculum pseudo labeling–
A hardness-aware curriculum learning framework that dynamically selects pseudo-labeled samples based on their difficulty, progressing from easy to complex examples is proposed, and both multi-stage and adaptive curriculum strategies to replace fixed-threshold filtering with more flexible, hardness-aware mechanisms are introduced.
- –
- –
- –
- –
- –
- CLIP-MT: Multi-Modal Knowledge-Driven Adaptive Scale Feature Allocation for Multi-Task Dense Prediction–
This work introduces a novel CLIP-Guided Global Feature Enhancer (CGGF), which leverages aligned text-image information to augment object-level representations through a dual-path feature fusion architecture, and designs an Adaptive Scale Selection Gate (ASSG), a learnable gating mechanism that dynamically selects high- or low-scale features based on task-specific demands.
- –
- –
- –
- Learning Multiple Representations with Inconsistency-Guided Detail Regularization for Mask-Guided Matting–
A novel auxiliary learning framework for mask-guided matting models is proposed, incorporating three auxiliary tasks: semantic segmentation, edge detection, and background line detection besides matting, to learn different and effective representations from different types of data and annotations.
- Research on automatic information extraction and combination based on semantic association analysis–
Information automatic combination tool is used to combine the extracted information, and according to the target characteristics, the information with dependence, sequence and similarity relationship is associated together to conduct intelligence association and intelligence clustering, so as to provide the complete information related to the task required for the user's decision.
- Research on Task Scenario-oriented Information User Concern Mining Technology–
The mission-scenario-oriented information user concern mining technology is an important component of the mission-Scenario- oriented information product adaptive service technology, which can support the collection of operational user behavior and the analysis of behavior rules in the interactive process of operational information.
- –
- –
- –
- –
- Data Quality Management in Institutional Research Output Data Center–
Institutional research output data center will store normative and convinced scholar’s research outputData will effectively support dynamic presentation of research output, reveal institutional academic publication in multiple dimensions, advance open access, and provide data support for subject evaluation and discipline development.
- Modeling long range relations by feature translation–
The proposed additional module containing a process called feature translation, to gather long range information at early stages has connection with dilated convolution and is more efficient and significantly improves performance in pose estimation.
- –
- –
- –
- –
- Fast Visual Object Tracking Using Convolutional Filters–
This paper theoretically explains the reason why synthetic exact filters based methods work well, proposes a novel visual object tracking algorithm based on convolutional filters, which are trained only by training images without labels and utilizes spatial context information to improve robustness of the tracking system.
- –
- –
- –
- –
- An application of partial least square model in the research of carrying capacity of water resources–
- Automated Defect Classification in Quality Management of TFT-LCD–
This paper presents a new method for automated defect classification (ADC), SIFT and LAB features are introduced to describe the defect regions, then a new kind of SVM-based Adaboost named DABsvm is proposed to classify the defect samples.
- –
- Image Classification Based on Weight Adjustment before Feature Pooling–
This paper proposes a new scheme that makes full use of spatial structure information to readjust their relative weights red and thus give some descriptors more chances to appear in the final feature vector more than others.
- Perception Preserving Projections–
Perception Preserving Projections (PPP) is proposed, which incorporates domain-specific feature extractor into the standard PCA formulation for the projection learning procedure, and enables PPP to make more sensible projections for feature based perception systems while retaining the simplicity and unsupervised manner of PCA.
- Neighborhood Structure Preserving Ridge Regression for Dimensionality Reduction–
A new algorithm for affinity graph construction based on nonnegative least squares and use affinity graph to capture the neighborhood geometric structure information and design an efficient model selection scheme for the optimal parameter estimation, which balances the tradeoff between the global and neighborhood structures.
- –
- The Watermark Based Anti-counterfeiting Information System for Forensic Digital Imaging–
The system, which can ensure the evidence effectiveness of digital images, consists of encrypting and anti-counterfeiting subsystem, information management subsystem and verification subsystem for encrypting information.
- –
- –
- Steganalysis Based on JPEG Binary Compressed Stream–
A method of steganalysis based on JPEG binary compressed stream which utilizes the changes of binary compressed data stream caused by the hidden messages, and uses Multiple Embedding Method (MEM) to draw a feature vector, which is used to train the One Class Support Vector Machine (OC-SVM).
- An improved Ant Colony Clustering Algorithm Using Information Entropy–
The paper intends to propose an improved entropy based ant colony clustering algorithm mainly for solving that problem through cluster splitting, merging and handling of isolated objects.
- –
- Improved Algorithm for Generating Two-Dimensional Body-Fitted Grid–
The grids generated by the improved TM-WEI method at the boundaries and the interior have better orthogonality than TM method and WEI Wenli method with higher accuracy of solution and better convenience of generation.
- MANIA: A Gene Network Reverse Algorithm for Compounds Mode-of-Action and Genes Interactions Inference–
A novel inference algorithm based on linear ordinary differential equations is proposed, based on high-throughput gene expression data, that can infer gene-gene interactions from steady state expression profiles and mode-of-action of the components that can trigger changes in the system.
- Blind Image Watermark Analysis Using Feature Fusion and Neural Network Classifier–
A general blind watermarking analysis scheme to recognize whether images are watermarked no matter what kind of watermark embedding schemes are used, using multiscale feature fusion and RBF neural networks.
- Penetration Ability of Aircraft to The Radar Net in Jamming Condition–
This paper analyzes the signal noise ratio (SNR) received by radar under the jamming condition, builds the detection probability model of radar net to the penetration aircraft and the survival probabilities model of the Penetration aircraft, which can provide the reference for the penetration combat to the radar net under theJamming condition.
- –
- A Double-SVM Classification System for Single andMultiple-Subcell ular–
The SVM Classification System toclassify protein sequences into develop a novel Double-SVM classification System for twocategories according to their primary structures: proteins predicting the subcellular localization sites of theproteins.
- THE MODELLING OF WEIGHTED COMPLEX NETWORKS–
A new model that incorporates the network topology and the weights' dynamical evolutions and calculates analytically the distributions of both degree and strength and found that all these distributions show scale-free behavior, as confirmed in many real networks.
- Image Fakery and Neural Network Based Detection–
A fake image detection scheme is proposed, which uses radial basis function (RBF) neural network as a detector to make a binary decision on whether an image is fake or real.
- Study of Information Stealth Techniques and Anti-jamming Ability of Stealth Communication–
The definition, basic composition, and its anti-jamming ability in stealth communication are introduced and basic composition is analyzed.
- –
- –
- –
- Some statistical analysis of the recall probability of parallel intraconnected bidirectional associative memory–
Two MIBAM with the equal total number of neurons have the equal recalling probability form pairs of stored pattern pairs ifm is not too large, i.
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 2026-10-15. 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.