Article
Computer Science, Artificial Intelligence
Aboozar Ghaffari, Mahdi Kafaee, Vahid Abolghasemi
Summary: In this paper, a novel inexpensive sparse non-negative reconstruction method, SnSA, is proposed to address the challenges in simultaneously achieving sparsity and non-negativity in image data representation. By introducing a non-negativity penalty term and a novel thresholding strategy, the SnSA algorithm avoids hard zeroing of negative samples to attain a balanced solution maximizing sparsity and minimizing the reconstruction error, achieving higher classification performance on synthetic signals and established image databases compared to state-of-the-art techniques.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Information Systems
Shicheng Yang, Ying Wen, Lianghua He, Mengchu Zhou, Abdullah Abusorrah
Summary: This work introduces a sparse individual low-rank component-based representation (SILR) method that effectively addresses the impact of undersampled training datasets and same intrasubject variations on classification performance by applying l(2)-norm constraint to intrasubject coefficients.
IEEE INTERNET OF THINGS JOURNAL
(2021)
Article
Computer Science, Information Systems
Xuqin Wei, Yun Shi, Weiyin Gong, Yanyun Guan
Summary: This paper introduces a novel image classification algorithm that uses an improved image representation method to generate virtual samples and designs a weight fusion scheme. The proposed algorithm improves the accuracy of image classification.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Metallurgy & Metallurgical Engineering
Tang De-yan, Zhou Si-wang, Luo Meng-ru, Chen Hao-wen, Tang Hui
Summary: In this study, a novel face recognition classifier called DSP is proposed, which achieves a better recognition rate than other classifiers in various situations and gradually removes most uncorrelated samples through iterations.
JOURNAL OF CENTRAL SOUTH UNIVERSITY
(2022)
Article
Computer Science, Information Systems
Xianzhong Long, Zhiyi Zhang, Yun Li
Summary: A new singular value decomposition based classification (SVDC) method is proposed for face recognition, which achieves better recognition performance and robustness in experiments.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Mathematics
He-Feng Yin, Xiao-Jun Wu, Cong Hu, Xiaoning Song
Summary: This study proposes a compact feature representation method using second-order image gradient orientations, achieving better performance in facial image classification tasks by applying linear complex principal component analysis (PCA) and collaborative-representation-based classification algorithm.
Article
Computer Science, Information Systems
Shicheng Yang, Ying Wen, Lianghua He, MengChu Zhou
Summary: This study proposes a novel method called sparse common feature-based representation (SCFR) to address the issue of undersampled face recognition encountered in IoT applications, providing better performance without the time-consuming training required by deep learning models.
IEEE INTERNET OF THINGS JOURNAL
(2021)
Article
Computer Science, Artificial Intelligence
Chao Zhang, Huaxiong Li, Chunlin Chen, Yuhua Qian, Xianzhong Zhou
Summary: This paper proposes a novel Enhanced Group Sparse regularized Nonconvex Regression (EGSNR) method for robust face recognition. By introducing an upper bounded nonconvex function to replace l(1)-norm for sparsity and combining gamma-norm and matrix gamma-norm to capture the characteristics of complex errors, EGSNR achieves more discriminative representation coefficients. Experimental results demonstrate that the proposed EGSNR outperforms the state-of-the-art regression based methods for robust face recognition.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Jian-Xun Mi, Qiang Huang, Li-Fang Zhou
Summary: This paper proposes a novel method called local spatial continuity steered sparse representation (LSCSR) for face recognition, which takes advantage of the local spatial continuity of occlusion to calculate the occlusion support map. The weighted sparse coding framework is then used for face representation and classification, showing effectiveness and robustness in the face recognition against occlusion and other variations.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Xiaojin Fan, Mengmeng Liao, Jingfeng Xue, Hao Wu, Lei Jin, Jian Zhao, Liehuang Zhu
Summary: In this paper, a Joint Coupled Representation and Homogeneous Reconstruction (JCRHR) method is proposed for multi-resolution small sample face recognition. The method improves the coherent representation of coding coefficients and the reconstruction effect of samples at different resolutions by introducing an analysis dictionary, a synthetic dictionary, and a coherence enhancement term. Experimental results demonstrate that the proposed JCRHR method outperforms existing methods on several small sample face databases.
Article
Computer Science, Information Systems
Susmini Indriani Lestariningati, Andriyan Bayu Suksmono, Ian Joseph Matheus Edward, Koredianto Usman
Summary: Sparse Representation-based Classification (SRC) is a reliable Face Recognition technique. This paper introduces a modified algorithm called GCR-SRC, which extends the coherency between test samples and identified training samples to improve recognition accuracy. The proposed method also reduces dimensionality and computational cost through random projection.
Article
Computer Science, Information Systems
Jing Li, Xiao Wei, Fengpin Wang, Jinjia Wang
Summary: A novel Inertial Proximal Gradient Method (IPGM) is proposed for convolutional dictionary learning, optimizing data fidelity and sparsity terms with inertial force, deriving new derivative formulas, using gradient descent steps, and proving convergence in a backtracking case. Simulation results show superior performance compared to other methods with the same structure.
Article
Engineering, Electrical & Electronic
HanQin Cai, Jian-Feng Cai, Tianming Wang, Guojian Yin
Summary: This paper investigates the robust recovery problem for spectrally sparse signals and proposes a highly efficient nonconvex algorithm called ASAP. By utilizing the low-rank property of the Hankel matrix and employing fast computations with structured matrices, ASAP achieves high computational efficiency and low space complexity for robust recovery of corrupted low-rank matrices. The theoretical recovery guarantee with a linear convergence rate and empirical performance comparisons demonstrate the advantages of ASAP in terms of computational efficiency and robustness.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Yuan Sun, Zhenwen Ren, Chao Yang, Quansen Sun, Liwan Chen, Yanglong Ou
Summary: Image set classification has gained extensive attention due to its ability to overcome various variations, and the point-to-point distance-based methods have achieved promising performance. However, these methods fail to fully utilize the discrimination information between different gallery sets and assume the equal importance of all sets. Additionally, they often have high computational cost. To address these issues, we propose a novel method called SLSDL, which incorporates a self-weighted strategy and latent sparse normalization to enhance discrimination and reduce complexity. Experimental results demonstrate that SLSDL outperforms state-of-the-art competitors.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Operations Research & Management Science
Wen Huang, Ke Wei
Summary: This paper discusses the problem of minimizing a differentiable function and a nonsmooth function on a Riemannian manifold. Various versions of Riemannian proximal gradient methods have been proposed to solve this problem. However, their convergence analyses require exact solutions to the Riemannian proximal mapping, which is costly or impractical. In this paper, an inexact Riemannian proximal gradient method is studied, and it is proven that accurate solutions to the proximal mapping guarantee global convergence and local convergence rate based on the Riemannian Kurdyka-Lojasiewicz property. Practical conditions for the accuracy of solving the Riemannian proximal mapping are also provided. Experimental results on sparse principal component analysis validate the proposed practical conditions.
COMPUTATIONAL OPTIMIZATION AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Mingjie He, Jie Zhang, Shiguang Shan, Xiao Liu, Zhongqin Wu, Xilin Chen
Summary: In this paper, a novel method for simulating occlusion by dropping the activations of a group of neurons is proposed, along with an attention module to improve the contributions of non-occluded regions. Experimental results show that the proposed method achieves significant improvements in the robustness and accuracy of face recognition.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Information Systems
Cheng Wang, Bingpeng Ma, Hong Chang, Shiguang Shan, Xilin Chen
Summary: This paper introduces a novel Bi-directional Task-Consistent Learning (BTCL) person search framework, which includes a Target-Specific Detector (TSD) and a re-ID model with Dynamic Adaptive Learning Structure (DALS) to meet the consistency needs between the detection and re-ID stages. Experimental results show that the framework achieves state-of-the-art performance on two widely-used person search datasets.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Artificial Intelligence
Shikang Yu, Hu Han, Shiguang Shan, Xilin Chen
Summary: This paper proposes a novel semi-supervised cross-modality synthesis method (CMOS-GAN) that can leverage both paired and unpaired face images to learn a robust cross-modality synthesis model. Experimental results show that the proposed method outperforms the state-of-the-art in three cross-modality face synthesis tasks.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Wenbin Wang, Ruiping Wang, Shiguang Shan, Xilin Chen
Summary: This study proposes a method for modeling human perception based on scene graphs, which generates a hierarchically constructed scene graph to present the primary content first and then the secondary content on demand. The method represents the scene with a hierarchical structure and utilizes a hierarchical contextual propagation module to model the structural information. It further highlights the key relationships in the scene graph through relationship re-ranking. Experimental results demonstrate that the method achieves state-of-the-art performance in scene graph generation and excels at mining image-specific relationships.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Yong Li, Shiguang Shan
Summary: In this study, a Meta Auxiliary Learning method (MAL) is proposed to automatically select highly related facial expression (FE) samples for better facial action unit (AU) detection performance. Experimental results demonstrate that MAL consistently improves AU detection performance compared with state-of-the-art multi-task and auxiliary learning methods.
IEEE TRANSACTIONS ON AFFECTIVE COMPUTING
(2023)
Proceedings Paper
Computer Science, Cybernetics
Fei Chang, Jiabei Zeng, Qiaoyun Liu, Shiguang Shan
Summary: Analyzing gaze patterns is crucial for understanding communication. Current studies have mainly focused on detecting single patterns, such as mutual gaze or shared attention. This work redefines five static gaze patterns that cover all statuses during dyadic communication and proposes a network to recognize these mutually exclusive patterns from images. Experimental results demonstrate that our method outperforms other solutions on a benchmark for gaze pattern recognition. Our method also achieves state-of-the-art performance on two other single gaze pattern recognition tasks. Analysis of gaze patterns in preschool children confirms findings in psychology.
ACM SYMPOSIUM ON EYE TRACKING RESEARCH & APPLICATIONS, ETRA 2023
(2023)
Article
Computer Science, Artificial Intelligence
Yong Li, Shiguang Shan
Summary: In this paper, the problem of insufficient AU annotations in facial action unit (AU) detection is addressed by learning AU representations from unlabelled facial videos through contrastive learning. The proposed method achieves frame-wise discriminative AU representations within a short video clip and consistent AU representations for facial frames with analogous AUs sampled from different identities. Experimental results demonstrate the discriminative power of the learned AU representation for AU detection.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Difei Gao, Ruiping Wang, Shiguang Shan, Xilin Chen
Summary: Inferring on visual facts and commonsense is crucial for an advanced VQA system. This requires models to surpass literal understanding of commonsense and fully ground it to the visual world. To evaluate this ability comprehensively, we introduce a VQA benchmark called CRIC, which includes new types of questions and an evaluation metric integrating correctness of answering and commonsense grounding. We also propose an automatic algorithm to generate question samples from scene and knowledge graphs, and analyze VQA models on the CRIC dataset. Experimental results demonstrate the challenge of grounding commonsense to image regions and joint reasoning on vision and commonsense. The dataset is available at https://cricvqa.github.io.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Ruibing Hou, Hong Chang, Bingpeng Ma, Shiguang Shan, Xilin Chen
Summary: In this study, we propose Dual Compensation Residual Networks to tackle the challenge of learning generalizable representation and classifier for class-imbalanced data in data-driven deep models. We introduce dual Feature Compensation Module (FCM) and Logit Compensation Module (LCM) to alleviate overfitting, and Residual Balanced Multi-Proxies Classifier (RBMC) to mitigate underfitting. Experimental results demonstrate the effectiveness of our approach.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Yude Wang, Jie Zhang, Meina Kan, Shiguang Shan, Xilin Chen
Summary: Scribble-supervised semantic segmentation is a low-cost weakly supervised technique that diffuses the labeled region of scribble to narrow the supervision gap. However, there is an annotation bias and label preference that affects the model training. To address this, BLPSeg is proposed to balance the label preference and design a local aggregation module to alleviate its impact.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Xuyang Guo, Meina Kan, Tianle Chen, Shiguang Shan
Summary: This study proposes an efficient and controllable method for continuous and fine hair editing using sliding bars, while also supporting discrete editing through reference photos and user-painted masks.
COMPUTER VISION - ECCV 2022, PT XV
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Zheng Yuan, Jie Zhang, Shiguang Shan
Summary: This paper introduces a novel architecture called Adaptive Image Transformation Learner (AITL) to improve the transferability of adversarial examples. Unlike existing works that use fixed combinational transformations, AITL improves the attack success rates by adaptively selecting the most effective combination of image transformations.
COMPUTER VISION - ECCV 2022, PT V
(2022)
Article
Computer Science, Artificial Intelligence
Nan Kang, Hong Chang, Bingpeng Ma, Shiguang Shan
Summary: This work aims to improve the feature extractor and classifier for long-tailed recognition through contrastive pretraining and feature normalization. It proposes a new balanced contrastive loss and fast contrastive initialization for better pretraining performance. It also introduces a novel generalized normalization classifier that outperforms traditional classifiers. The unified framework achieves competitive performance on long-tailed recognition benchmarks while maintaining high efficiency.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Yong Li, Lingjie Lao, Zhen Cui, Shiguang Shan, Jian Yang
Summary: This study proposes a method called GraphJigsaw for cartoon face recognition, which utilizes jigsaw puzzles and graph convolutional networks. The proposed method achieves better accuracy compared to other existing methods without requiring manual annotations or additional computational burden during inference.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Xinqian Gu, Hong Chang, Bingpeng Ma, Shiguang Shan
Summary: The proposed Motion Feature Aggregation (MFA) method efficiently models and aggregates motion information in the feature map level for video-based re-identification (re-id), consisting of coarse-grained and fine-grained motion learning modules that can model motion information from different granularities and are complementary to each other.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Rui Lv, Dingheng Wang, Jiangbin Zheng, Zhao-Xu Yang
Summary: In this paper, the authors investigate tensor decomposition for neural network compression. They analyze the convergence and precision of tensor mapping theory, validate the rationality of tensor mapping and its superiority over traditional tensor approximation based on the Lottery Ticket Hypothesis. They propose an efficient method called 3D-KCPNet to compress 3D convolutional neural networks using the Kronecker canonical polyadic (KCP) tensor decomposition. Experimental results show that 3D-KCPNet achieves higher accuracy compared to the original baseline model and the corresponding tensor approximation model.
Article
Computer Science, Artificial Intelligence
Xiangkun He, Zhongxu Hu, Haohan Yang, Chen Lv
Summary: In this paper, a novel constrained multi-objective reinforcement learning algorithm is proposed for personalized end-to-end robotic control with continuous actions. The approach trains a single model using constraint design and a comprehensive index to achieve optimal policies based on user-specified preferences.
Article
Computer Science, Artificial Intelligence
Zhijian Zhuo, Bilian Chen, Shenbao Yu, Langcai Cao
Summary: In this paper, a novel method called Expansion with Contraction Method for Overlapping Community Detection (ECOCD) is proposed, which utilizes non-negative matrix factorization to obtain disjoint communities and applies expansion and contraction processes to adjust the degree of overlap. ECOCD is applicable to various networks with different properties and achieves high-quality overlapping community detection.
Article
Computer Science, Artificial Intelligence
Yizhe Zhu, Chunhui Zhang, Jialin Gao, Xin Sun, Zihan Rui, Xi Zhou
Summary: In this work, the authors propose a Contrastive Spatio-Temporal Distilling (CSTD) approach to improve the detection of high-compressed deepfake videos. The approach leverages spatial-frequency cues and temporal-contrastive alignment to fully exploit spatiotemporal inconsistency information.
Review
Computer Science, Artificial Intelligence
Laijin Meng, Xinghao Jiang, Tanfeng Sun
Summary: This paper provides a review of coverless steganographic algorithms, including the development process, known contributions, and general issues in image and video algorithms. It also discusses the security of coverless steganography from theoretical analysis to actual investigation for the first time.
Article
Computer Science, Artificial Intelligence
Yajie Bao, Tianwei Xing, Xun Chen
Summary: Visual question answering requires processing multi-modal information and effective reasoning. Neural-symbolic learning is a promising method, but current approaches lack uncertainty handling and can only provide a single answer. To address this, we propose a confidence based neural-symbolic approach that evaluates NN inferences and conducts reasoning based on confidence.
Article
Computer Science, Artificial Intelligence
Anh H. Vo, Bao T. Nguyen
Summary: Interior style classification is an interesting problem with potential applications in both commercial and academic domains. This project proposes a method named ISC-DeIT, which combines data-efficient image transformer architectures and knowledge distillation, to address the interior style classification problem. Experimental results demonstrate a significant improvement in predictive accuracy compared to other state-of-the-art methods.
Article
Computer Science, Artificial Intelligence
Shashank Kotyan, Danilo Vasconcellos Vargas
Summary: This article introduces a novel augmentation technique called Dynamic Scanning Augmentation to improve the accuracy and robustness of Vision Transformer (ViT). The technique leverages dynamic input sequences to adaptively focus on different patches, resulting in significant changes in ViT's attention mechanism. Experimental results demonstrate that Dynamic Scanning Augmentation outperforms ViT in terms of both robustness to adversarial attacks and accuracy against natural images.
Article
Computer Science, Artificial Intelligence
Hiba Alqasir, Damien Muselet, Christophe Ducottet
Summary: The article proposes a solution to improve the learning process of a classification network by providing shape priors, reducing the need for annotated data. The solution is tested on cross-domain digit classification tasks and a video surveillance application.
Article
Computer Science, Artificial Intelligence
Dexiu Ma, Mei Liu, Mingsheng Shang
Summary: This paper proposes a method using neural dynamics solvers to solve infinity-norm optimization problems. Two improved solvers are constructed and their effectiveness and superiority are demonstrated through theoretical analysis and simulation experiments.
Article
Computer Science, Artificial Intelligence
Francesco Gregoretti, Giovanni Pezzulo, Domenico Maisto
Summary: Active Inference is a computational framework that uses probabilistic inference and variational free energy minimization to describe perception, planning, and action. cpp-AIF is a header-only C++ library that provides a powerful tool for implementing Active Inference for Partially Observable Markov Decision Processes through multi-core computing. It is cross-platform and improves performance, memory management, and usability compared to existing software.
Article
Computer Science, Artificial Intelligence
Zelin Ying, Dawei Cheng, Cen Chen, Xiang Li, Peng Zhu, Yifeng Luo, Yuqi Liang
Summary: This paper proposes a novel stock market trends prediction framework called SMART, which includes a self-supervised stock technical data sequence embedding model S3E. By training with multiple self-supervised auxiliary tasks, the model encodes stock technical data sequences into embeddings and uses the learned sequence embeddings for predicting stock market trends. Extensive experiments on China A-Shares market and NASDAQ market prove the high effectiveness of our model in stock market trends prediction, and its effectiveness is further validated in real-world applications in a leading financial service provider in China.
Article
Computer Science, Artificial Intelligence
Hao Li, Hao Jiang, Dongsheng Ye, Qiang Wang, Liang Du, Yuanyuan Zeng, Liu Yuan, Yingxue Wang, C. Chen
Summary: DHGAT1, a dynamic hyperbolic graph attention network, utilizes hyperbolic metric properties to embed dynamic graphs. It employs a spatiotemporal self-attention mechanism and weighted node representations, resulting in excellent performance in link prediction tasks.
Article
Computer Science, Artificial Intelligence
Jiehui Huang, Zhenchao Tang, Xuedong He, Jun Zhou, Defeng Zhou, Calvin Yu-Chian Chen
Summary: This study proposes a progressive learning multi-scale feature blending model for image deraining tasks. The model utilizes detail dilation and texture extraction to improve the restoration of rainy images. Experimental results show that the model achieves near state-of-the-art performance in rain removal tasks and exhibits better rain removal realism.
Article
Computer Science, Artificial Intelligence
Lizhi Liu, Zilin Gao, Yinhe Wang, Yongfu Li
Summary: This paper proposes a novel discrete-time interconnected model for depicting complex dynamical networks. The model consists of nodes and edges subsystems, which consider the dynamic characteristic of both nodes and edges. By designing control strategies and coupling modes, the stabilization and synchronization of the network are achieved. Simulation results demonstrate the effectiveness of the proposed methods.