模式识别与人工智能
Thursday, Apr. 3, 2025 Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2025 Vol.38 Issue.2, Published 2025-02-25

   
101 Image Inpainting Based on Global-Local Prior and Texture Details
XU Qijin, YE Hailiang, CAO Feilong, LIANG Jiye
Image inpainting is intended to fill in missing regions of an image using surrounding information. However, existing prior-based methods often struggle to balance global semantic consistency and local texture details. In this paper, a method for image inpainting based on global-local prior and texture details is proposed. Wavelet-Fourier convolution blocks are constructed by combining wavelet convolution and Fourier convolution to enhance the interaction between local and global features. Based on the above, a global-local learning-based prior is presented. A prior extractor composed of wavelet-Fourier convolution blocks is designed to simultaneously learn global and local priors. The prior extractor is applied to both damaged and complete images to obtain damaged priors and supervised priors. During the repair phase, the damaged image and the learned priors are input into two structurally similar repair branches. Both branches are constructed with wavelet-Fourier convolutions and can simultaneously extract and fuse global and local features. Finally, the outputs of the two branches are merged to generate the image with consistent semantic content and clear local details. Additionally, a high receptive field style loss is introduced to improve image style consistency at the semantic level. Experimental results show that the proposed method outperforms existing methods on multiple datasets.
2025 Vol. 38 (2): 101-115 [Abstract] ( 35 ) [HTML 1KB] [ PDF 4498KB] ( 13 )
116 Correspondence Calculation for Non-isometric 3D Shape Collection via Coupled Maps
YANG Jun, XUE Youzhong
To address the issues of low accuracy and poor generalization ability in existing non-isometric 3D shape collection correspondence calculation methods, a correspondence calculation method for non-isometric 3D shape collection via coupled maps is proposed. First, DiffusionNet is employed to directly extract initial features from the 3D shape, and thus discriminative feature descriptors are obtained. Then, functional maps matrix and point-to-point maps matrix are computed using these descriptors. Structural regularization constraints and softmax normalization are applied to both matrices, respectively, to obtain an optimal coupled maps matrix. Finally, a shape collection matching module based on a virtual template takes the initial model features as input and employs a point classifier constructed with the coupled maps to directly predict the correspondence between the shapes and the virtual templates. The final correspondence for the non-isometric shape collection is obtained through Gumbel-Sinkhorn normalization. Experimental results demonstrate that the proposed method effectively handles topological noise within non-isometric shapes, achieves low geodesic error in correspondence calculation, provides accurate results, and exhibits strong generalization ability.
2025 Vol. 38 (2): 116-131 [Abstract] ( 9 ) [HTML 1KB] [ PDF 4985KB] ( 6 )
132 Semantic-Based Prototype Optimization Method for Few-Shot Learning
LIU Yuanyuan, SHAO Mingwen, ZHANG Lixu, SHAO Xun
Semantic information can provide rich prior knowledge for few-shot learning. However, existing few-shot learning studies only superficially explore the combination of images and semantics, failing to fully utilize semantics to explore class features. Consequently, the model performance is limited. To address this issue, a semantic-based prototype optimization method for few-shot learning(SBPO) is proposed. First, SBPO employs channel-wise semantic prompts to guide the model in extracting visual features while progressively optimizing class prototypes. Second, a multi-modal margin loss is designed to integrate inter-class correlations in both visual and semantic dimensions with the loss function, thereby constraining the model to enhance the distinctiveness of class prototypes. Finally, through a two-stage fine-tuning process, the model can fully leverage semantic knowledge to optimize class prototypes, thereby improving classification accuracy. Experiments on four benchmark datasets demonstrate that SBPO significantly outperforms baseline methods.
2025 Vol. 38 (2): 132-142 [Abstract] ( 13 ) [HTML 1KB] [ PDF 1176KB] ( 11 )
143 Two-Layer Network Based Epidemic Model with Network Formal Context
FAN Min, CHEN Rui, LI Jinhai
Two-layer network based epidemic models are one of the hot topics in complex network dynamics. However, existing studies overlook the impact of epidemic awareness and behavior on epidemic transmission. As a result, when there are significant differences in individual prevention behaviors, these models fail to accurately reflect real-world disease spread. To address this issue, a two-layer network based epidemic model is proposed by integrating formal concept analysis with the microscopic Markov chain approach(MMCA) from the perspective of behavioral pattern recognition. First, the two-layer network formal context, network concepts and their characteristic parameters are defined, and a bridge between formal concept analysis and epidemic models is established. This method not only describes the concepts and characteristic parameters corresponding to behavioral patterns in the two-layer network, but also defines a decay factor to further facilitate the integration of information across layers using MMCA. Second, the influence of mass media and policy interventions on information diffusion is taken into account, the mass media function and the MMCA model are improved, and the epidemic outbreak threshold is derived. Finally, simulation experiments are conducted to analyze the impact of several key parameters on epidemic spread scale and threshold.
2025 Vol. 38 (2): 143-163 [Abstract] ( 19 ) [HTML 1KB] [ PDF 1037KB] ( 10 )
164 Human Pose Estimation Based on Knowledge Distillation and Dynamic Region Refinement
WEI Longsheng, FU Xingpeng, LI Tangqiang, HUANG Haoyu
Human pose estimation methods are categorized into coordinate regression-based methods and heatmap-based methods. Coordinate regression-based methods are characterized by slightly faster inference speed but slightly lower accuracy, while heatmap-based methods can achieve precise localization at the cost of higher computational and storage overhead. Therefore, a human pose estimation method based on knowledge distillation and dynamic region refinement is proposed. First, the information from the heatmap model is transferred to the regression model through feature distillation and pose distillation. Then, the features extracted by multi-layer Transformer are selected to generate initial pose estimation in the coarse stage, and the image features that need to be refined are selected based on the scores from a quality predictor. Finally, in the refinement stage, fine-grained representations or refined features, are established in the regions related to some keypoints according to the correlation between keypoints and image regions, achieving human pose refinement. Experiments on COCO and COCO-WholeBody datasets demonstrate that the proposed method can accurately locate keypoints and achieve accurate human pose estimation.
2025 Vol. 38 (2): 164-176 [Abstract] ( 15 ) [HTML 1KB] [ PDF 1313KB] ( 8 )
177 Hessian Aided Probabilistic Policy Gradient Method
HU Lei, LI Yongqiang, FENG Yu, FENG Yuanjing
Policy gradient methods in reinforcement learning are widely applied to continuous decision-making problems due to their generality. However, their practical performance is consistently constrained by low sample utilization caused by high gradient variance. In this paper, a Hessian-aided probabilistic policy gradient method(HAPPG) is proposed, and a bimodal gradient estimation mechanism is designed based on the probabilistic gradient estimator. Historical momentum is added to large-batch-size estimation to restrict optimization fluctuation of gradient descent, and variance-reduced estimation based on the Hessian-aided technique is constructed by introducing the second-order curvature information of the policy parameters into the small-batch-size estimation. Theoretical analysis demonstrates that HAPPG achieves an O($\epsilon$-3) sample complexity under non-convex optimization conditions, attaining the best convergence rate among the existing methods. Experimental results validate its superior performance across multiple benchmark control tasks. Furthermore, the Hessian-aided probabilistic policy gradient estimator is combined with the proximal policy optimization(PPO) by embedding the adaptive learning rate mechanism of Adam optimizer, resulting in HAP-PPO. HAP-PPO outperforms PPO, and the designed gradient estimator can be applied to further enhance mainstream reinforcement learning algorithms.
2025 Vol. 38 (2): 177-191 [Abstract] ( 19 ) [HTML 1KB] [ PDF 1138KB] ( 10 )
192
2025 Vol. 38 (2): 192-192 [Abstract] ( 20 ) [HTML 1KB] [ PDF 140KB] ( 8 )
模式识别与人工智能
 

Supervised by
China Association for Science and Technology
Sponsored by
Chinese Association of Automation
NationalResearchCenter for Intelligent Computing System
Institute of Intelligent Machines, Chinese Academy of Sciences
Published by
Science Press
 
Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence
Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn
Supported by Beijing Magtech  Email:support@magtech.com.cn