检索结果-内蒙古大学图书馆

Heterogeneous window Transformer for image denoising

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Tian, Chunwei Zheng, Menghua Lin, Chia-Wen Li, Zhiwu Zhang, David School of Software Northwestern Polytechnical University National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology Xi’an710129 China School of Software Northwestern Polytechnical University Xi’an710129 China Department of Electrical Engineering the Institute of Communications Engineering National Tsing Hua University 300 Taiwan School of Elector-Mechanical Engineering Xidian University Xi’an710071 China Shenzhen518172 China Shenzhen Institute of Artificial Intelligence and Robotics for Society Shenzhen518172 China

Deep networks can usually depend on extracting more structural information to improve denoising results. However, they may ignore correlation between pixels from an image to pursue better denoising performance. Window Transformer can use long- and short-distance modeling to interact pixels to address mentioned problem. To make a tradeoff between distance modeling and denoising time, we propose a heterogeneous window Transformer (HWformer) for image denoising. HWformer first designs heterogeneous global windows to capture global context information for improving denoising effects. To build a bridge between long and short-distance modeling, global windows are horizontally and vertically shifted to facilitate diversified information without increasing denoising time. To prevent the information loss phenomenon of independent patches, sparse idea is guided a feed-forward network to extract local information of neighboring patches. The proposed HWformer only takes 30% of popular Restormer in terms of denoising time. Its codes can be obtained at https://***/hellloxiaotian/HWformer. © 2024, CC BY-NC-ND.

关键词： Image denoising

HyperDefect-YOLO: Enhance YOLO with HyperGraph Computation for Industrial Defect Detection

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zuo, Zuo Dong, Jiahao Gao, Yue Wu, Zongze National Key Laboratory of Human-Machine Hybrid Augmented Intelligence National Engineering Research Center for Visual Information and Applications Institute of Artificial Intelligence and Robotics Xi’an Jiaotong University Xi’an710049 China China The School of Software BNRist THUIBCS BLBCI Ts-inghua University Beijing100084 China College of Mechatronics and Control Engineering Shenzhen University Shenzhen China Institute of Artificial Intelligence and Robotics Xi’an Jiaotong University Xi’an710049 China

In the manufacturing industry, defect detection is an essential but challenging task aiming to detect defects generated in the process of production. Though traditional YOLO models presents a good performance in defect detection, they still have limitations in capturing high-order feature interrelationships, which hurdles defect detection in the complex scenarios and across the scales. To this end, we introduce hypergraph computation into YOLO framework, dubbed HyperDefect-YOLO (HD-YOLO), to improve representative ability and semantic exploitation. HD-YOLO consists of Defect Aware Module (DAM) and Mixed Graph Network (MGNet) in the backbone, which specialize for perception and extraction of defect features. To effectively aggregate multi-scale features, we propose HyperGraph Aggregation Network (HGANet) which combines hypergraph and attention mechanism to aggregate multi-scale features. Cross-Scale Fusion (CSF) is proposed to adaptively fuse and handle features instead of simple concatenation and convolution. Finally, we propose Semantic Aware Module (SAM) in the neck to enhance semantic exploitation for accurately localizing defects with different sizes in the disturbed background. HD-YOLO undergoes rigorous evaluation on public HRIPCB and NEU-DET datasets with significant improvements compared to state-of-the-art methods. We also evaluate HD-YOLO on self-built MINILED dataset collected in real industrial scenarios to demonstrate the effectiveness of the proposed method. The source codes are at https://***/Jay-zzcoder/HD-YOLO. Copyright © 2024, The Authors. All rights reserved.

关键词： Aggregates

Unsupervised Video Anomaly Detection by Learning Consistency of Appearance and Motion

学校读者我要写书评

暂无评论

Unsupervised Video Anomaly Detection by Learning Consistency...

Chinese Automation Congress (CAC)

作者： Yanlong Li Zhenhao Ye Yang Wang Chi Zhang Yuehu Liu School of Software Engineering College of Artificial Intelligence National Key Laboratory of Human-Machine Hybrid Augmented Intelligence National Engineering Research Center for Visual Information and Applications Institute of Artificial Intelligence and Robotics Xian Jiaotong University Xi'an China

Unsupervised video anomaly detection aims to train the model on unlabeled videos containing only normal behavior and identify data instances with abnormal objects or behaviors as anomalies. Abnormal events are typically detected due to their unique appearance and motion features. Existing methods mainly rely on various appearance-motion feature fusion techniques to model the information from both modalities for anomaly detection. However, they overlook the semantic consistency between motion and appearance information, resulting in constrained model performance. To address this limitation, we directly model the relationship of semantic consistency between appearance and motion. Given the varying importance of different pixels, manually designing feature fusion methods to represent their consistency is impractical. So, we designed a two-stream feature fusion module based on cross-attention. This module constructs information fusion between appearance and motion semantics in normal videos, dynamically capturing the inherent appearance-motion semantic consistency. We further believe that, regardless of the perspective, the appearance-motion semantic consistency should be similar. Therefore, this paper introduces a new consistency loss, urging the model to fully model the semantic representation of consistency between appearance and motion in normal data, thereby identifying anomalies with lower consistency. Experimental results on various standard datasets and ablation studies have demonstrated the effectiveness of the proposed method.

关键词：

Pest Detection and Identification Guided by Feature Maps

学校读者我要写书评

暂无评论

Pest Detection and Identification Guided by Feature Maps

Workshops on Image Processing Theory, Tools and Applications, IPTA

作者： Miao Chen Yanan Chen Minghui Guo Jianji Wang National Key Laboratory of Human-Machine Hybrid Augmented Intelligence National Engineering Research Center for Visual Information and Applications and Institute of Artificial Intelligence and Robotics Xi’an Jiaotong University School of Software Engineering Xi’an Jiaotong University Xi’an China

Effective pest detection and identification are of great significance for agricultural activities, and morden machine learning methods, especially the deep neural network, undoubtedly provide convenient and effective ways for this issue. Most of the mainstream methods for pest detection require annotation of manual bounding box. However, the public pest datasets usually lack the labeling information. In this work, we propose a method for detecting pests with weak supervision. We obtain the key regions of the pest targets by deeply mining the information in the feature maps of the convolutional neural networks (CNNs). First, using the class activation mapping technique to generate activation maps of the target class during forward propagation of the model, and then obtaining the key regions of the target based on the hotspots extracted from the activation maps. In addition, our proposed method can be deployed as a general strategy to other CNNs, and experimental results demonstrate that our method can steadily improve the classification accuracy of the model. By applying our method to the single Resnet50, we can achieve 74.27% classification accuracy on the publicly accessible IP102 dataset. Another attribution of this paper is that we propose a lightweight approach to find solutions for pest detection in agricultural automation applications.

关键词：

Hypergraph Foundation Model

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Feng, Yifan Liu, Shiquan Han, Xiangmin Du, Shaoyi Wu, Zongze Hu, Han Gao, Yue School of Software BNRist THUIBCS BLBCI Tsinghua University Beijing100084 China Institute of Artificial Intelligence and Robotics College of Artificial Intelligence Xi’an Jiaotong University Xi’an710049 China College of Mechatronics and Control Engineering Shenzhen University Guangdong Shenzhen510006 China Beijing Institute of Technology Beijing100811 China

Hypergraph neural networks (HGNNs) effectively model complex high-order relationships in domains like protein interactions and social networks by connecting multiple vertices through hyperedges, enhancing modeling capabilities, and reducing information loss. Developing foundation models for hypergraphs is challenging due to their distinct data, which includes both vertex features and intricate structural information. We present Hyper-FM, a Hypergraph Foundation Model for multi-domain knowledge extraction, featuring Hierarchical High-Order Neighbor Guided Vertex Knowledge Embedding for vertex feature representation and Hierarchical Multi-Hypergraph Guided Structural Knowledge Extraction for structural information. Additionally, we curate 10 text-attributed hypergraph datasets to advance research between HGNNs and LLMs. Experiments on these datasets show that Hyper-FM outperforms baseline methods by approximately 13.3%, validating our approach. Furthermore, we propose the first scaling law for hypergraph foundation models, demonstrating that increasing domain diversity significantly enhances performance, unlike merely augmenting vertex and hyperedge counts. This underscores the critical role of domain diversity in scaling hypergraph models. Copyright © 2025, The Authors. All rights reserved.

关键词： Neural network models

STD-NET: Search of Image Steganalytic Deep-learning Architecture via Hierarchical Tensor Decomposition

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Tan, Shunquan Li, Qiushi Li, Laiyuan Li, Bin Huang, Jiwu Shenzhen Institute of Artificial Intelligence and Robotics for Society China College of Computer Science and Software Engineering Shenzhen University China College of Electronic and Information Engineering Shenzhen University China

Recent studies shows that the majority of existing deep steganalysis models have a large amount of redundancy, which leads to a huge waste of storage and computing resources. The existing model compression method cannot flexibly compress the convolutional layer in residual shortcut block so that a satisfactory shrinking rate cannot be obtained. In this paper, we propose STD-NET, an unsupervised deep-learning architecture search approach via hierarchical tensor decomposition for image steganalysis. Our proposed strategy will not be restricted by various residual connections, since this strategy does not change the number of input and output channels of the convolution block. We propose a normalized distortion threshold to evaluate the sensitivity of each involved convolutional layer of the base model to guide STD-NET to compress target network in an efficient and unsupervised approach, and obtain two network structures of different shapes with low computation cost and similar performance compared with the original one. Extensive experiments have confirmed that, on one hand, our model can achieve comparable or even better detection performance in various steganalytic scenarios due to the great adaptivity of the obtained network architecture. On the other hand, the experimental results also demonstrate that our proposed strategy is more efficient and can remove more redundancy compared with previous steganalytic network compression methods. Copyright © 2022, The Authors. All rights reserved.

关键词： Redundancy

Towards Generalizable Multi-Object Tracking

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Qin, Zheng Wang, Le Zhou, Sanping Fu, Panpan Hua, Gang Tang, Wei National Key Laboratory of Human-Machine Hybrid Augmented Intelligence National Engineering Research Center for Visual Information and Applications Institute of Artificial Intelligence and Robotics Xi’an Jiaotong University China School of Software Engineering Xi’an Jiaotong University China Wormpex AI Research University of Illinois Chicago United States

Multi-Object Tracking (MOT) encompasses various tracking scenarios, each characterized by unique traits. Effective trackers should demonstrate a high degree of generalizability across diverse scenarios. However, existing trackers struggle to accommodate all aspects or necessitate hypothesis and experimentation to customize the association information (motion and/or appearance) for a given scenario, leading to narrowly tailored solutions with limited generalizability. In this paper, we investigate the factors that influence trackers’ generalization to different scenarios and concretize them into a set of tracking scenario attributes to guide the design of more generalizable trackers. Furthermore, we propose a "point-wise to instance-wise relation" framework for MOT, i.e., GeneralTrack, which can generalize across diverse scenarios while eliminating the need to balance motion and appearance. Thanks to its superior generalizability, our proposed GeneralTrack achieves state-of-the-art performance on multiple benchmarks and demonstrates the potential for domain generalization. https://***/qinzheng2000/*** Copyright © 2024, The Authors. All rights reserved.

关键词： Benchmarking

DynamicAvatars: Accurate Dynamic Facial Avatars Reconstruction and Precise Editing with Diffusion Models

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Qian, Yangyang Sun, Yuan Guo, Yu School of Software Engineering Xi’an Jiaotong University China National Key Laboratory of Human-Machine Hybrid Augmented Intelligence Xi’an Jiaotong University China National Engineering Research Center for Visual Information and Applications Xi’an Jiaotong University China Institute of Artificial Intelligence and Robotics Xi’an Jiaotong University China

Generating and editing dynamic 3D head avatars are crucial tasks in virtual reality and film production. However, existing methods often suffer from facial distortions, inaccurate head movements, and limited fine-grained editing capabilities. To address these challenges, we present DynamicAvatars, a dynamic model that generates photorealistic, moving 3D head avatars from video clips and parameters associated with facial positions and expressions. Our approach enables precise editing through a novel prompt-based editing model, which integrates user-provided prompts with guiding parameters derived from large language models[26] (LLMs). To achieve this, we propose a dual-tracking framework based on Gaussian Splatting and introduce a prompt preprocessing module to enhance editing stability. By incorporating a specialized GAN algorithm and connecting it to our control module, which generates precise guiding parameters from LLMs, we successfully address the limitations of existing methods. Additionally, we develop a dynamic editing strategy that selectively utilizes specific training datasets to improve the efficiency and adaptability of the model for dynamic editing tasks. Copyright © 2024, The Authors. All rights reserved.

关键词： 3D reconstruction

Coordinate transformation and connection feature for Skeleton-based action recognition

学校读者我要写书评

暂无评论

Coordinate transformation and connection feature for Skeleto...

Chinese Automation Congress (CAC)

作者： Junling Zheng Wensong Chan Shaoyi Du Fei Zhao Zhiqiang Tian School of Software Engineering Xi’an Jiaotong University Xi’an China School of Software Engineering Institute of Artificial Intelligence and Robotics School of Mechanical Engineering

ISBN: (纸本)9781665426480

Graph structure is an important part of Graph convolutional networks (GCNs), which can reflect the connection between each nodes of non-Euclidean data. A connection feature between nodes is hidden in graph structure, which can provide additional spatial features that represent the relationship between human joints. However many GCNs-based methods ignore these spatial features. We put forward a connection feature extraction module, which can obtain implicit connection between human joints, and extract the implicit spatial features from the structural connection and implicit connection of human joints. In order to enhance the temporal representation, we propose a long-range frame-difference feature extraction module. Furthermore, we also propose a coordinate transformation module, which can map joint from Cartesian coordinates to spherical coordinates to extract more representative features. Experiments show that our method outperforms several advanced methods.

关键词： Automation Transforms Feature extraction