检索结果-内蒙古大学图书馆

IEEE/CVF International Conference on Computer Vision (ICCV)

作者： Lin, Kunyang Chen, Peihao Huang, Diwei Li, Thomas H. Tan, Mingkui Gan, Chuang South China Univ Technol Guangzhou Peoples R China Peking Univ Informat Technol R&D Innovat Ctr Beijing Peoples R China UMass Amherst Amherst MA USA MIT IBM Watson Lab Cambridge MA USA Minist Educ Key Lab Big Data & Intelligent Robot Beijing Peoples R China Peking Univ Shenzhen Grad Sch Beijing Peoples R China

ISBN: (纸本)9798350307184

Vision-and-language navigation (VLN) requires an embodied agent to navigate in realistic 3D environments using natural language instructions. Existing VLN methods suffer from training on small-scale environments or unreasonable path-instruction datasets, limiting the generalization to unseen environments. There are massive house tour videos on YouTube, providing abundant real navigation experiences and layout information. However, these videos have not been explored for VLN before. In this paper, we propose to learn an agent from these videos by creating a large-scale dataset which comprises reasonable path-instruction pairs from house tour videos and pre-training the agent on it. To achieve this, we have to tackle the challenges of automatically constructing path-instruction pairs and exploiting real layout knowledge from raw and unlabeled videos. To address these, we first leverage an entropy-based method to construct the nodes of a path trajectory. Then, we propose an action-aware generator for generating instructions from unlabeled trajectories. Last, we devise a trajectory judgment pretext task to encourage the agent to mine the layout knowledge. Experimental results show that our method achieves state-of-the-art performance on two popular benchmarks (R2R and REVERIE). Code is available at https://***/JeremyLinky/YouTube-VLN

关键词：

来源：评论

学校读者我要写书评

暂无评论

Cascaded Hierarchical Context-Aware Vehicle Re-Identification

Cascaded Hierarchical Context-Aware Vehicle Re-Identificatio...

引用

International Joint Conference on Neural Networks (IJCNN)

作者： Mo, Wancheng Lv, Jianming South China Univ Technol Sch Comp Sci & Engn Guangzhou Guangdong Peoples R China South China Univ Technol Key Lab Big Data & Intelligent Robot Guangzhou Guangdong Peoples R China

ISBN: (纸本)9780738133669

Vehicle Re-Identification (Re-ID) is a challenging task, which aims to match the surveillance images containing the same vehicle. Since vehicles of the same type tend to share very similar appearance, slight difference in local areas are usually critical in the vehicle Re-ID task. Recently, some fine-grained Re-ID algorithms have achieved superior performance by modeling the key areas with specific semantics such as windows, lights, car orientation, etc. However, such methods are labor-intensive to label the key areas for object detection. This work proposes a Cascaded Hierarchical Context-Aware scheme namely CHCA, which is free of fine-grained labeling, to adaptively extract the visual features of discriminative local areas based on surrounding hierarchical context information with a specially designed recursive cross-level attention mechanism. It does not require any additional supervision and is easy to be embedded in existing networks. Extensive experiments on three popular vehicle Re-ID benchmarks demonstrate the effectiveness of CHCA, which has competitive results with existing state-of-the-art methods applying fine-grained labels.

关键词： Visualization Surveillance Semantics Neural networks Object detection Benchmark testing Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Efficient Test-Time Model Adaptation without Forgetting 39

Efficient Test-Time Model Adaptation without Forgetting

引用

39th International Conference on Machine Learning (ICML)

作者： Niu, Shuaicheng Wu, Jiaxiang Zhang, Yifan Chen, Yaofo Zheng, Shijian Zhao, Peilin Tan, Mingkui Tencent AI Lab Shenzhen Peoples R China South China Univ Technol Sch Software Engn Guangzhou Guangdong Peoples R China Pazhou Lab Guangzhou Peoples R China Natl Univ Singapore Singapore Singapore Minist Educ Key Lab Big Data & Intelligent Robot Guangzhou Peoples R China

Test-time adaptation (TTA) seeks to tackle potential distribution shifts between training and testing data by adapting a given model w.r.t. any testing sample. This task is particularly important for deep models when the test environment changes frequently. Although some recent attempts have been made to handle this task, we still face two practical challenges: 1) existing methods have to perform backward computation for each test sample, resulting in unbearable prediction cost to many applications;2) while existing TTA solutions can significantly improve the test performance on out-of-distribution data, they often suffer from severe performance degradation on in-distribution data after TTA (known as catastrophic forgetting). In this paper, we point out that not all the test samples contribute equally to model adaptation, and high-entropy ones may lead to noisy gradients that could disrupt the model. Motivated by this, we propose an active sample selection criterion to identify reliable and non-redundant samples, on which the model is updated to minimize the entropy loss for test-time adaptation. Furthermore, to alleviate the forgetting issue, we introduce a Fisher regularizer to constrain important model parameters from drastic changes, where the Fisher importance is estimated from test samples with generated pseudo labels. Extensive experiments on CIFAR-10-C, ImageNet-C, and ImageNet-R verify the effectiveness of our proposed method. Code is available at https://***/mr-eggplant/EATA.

关键词：

来源：评论

学校读者我要写书评

暂无评论

LSU-NET: LIGHTWEIGHT AUTOMATIC ORGANS SEGMENTATION NETWORK FOR MEDICAL IMAGES

arXiv

引用

arXiv 2025年

作者： Ding, Yujie Teng, Shenghua Li, Zuoyong Chen, Xiao College of Electronic and Information Engineering Shandong University of Science and Technology Qingdao266590 China Fujian Provincial Key Laboratory of Information Processing and Intelligent Control School of Computer and Big Data Minjiang University Fuzhou350121 China

UNet and its variants have widespread applications in medical image segmentation. However, the substantial number of parameters and computational complexity of these models make them less suitable for use in clinical settings with limited computational resources. To address this limitation, we propose a novel Lightweight Shift U-Net (LSU-Net). We integrate the Light Conv Block and the Tokenized Shift Block in a lightweight manner, combining them with a dynamic weight multi-loss design for efficient dynamic weight allocation. The Light Conv Block effectively captures features with a low parameter count by combining standard convolutions with depthwise separable convolutions. The Tokenized Shift Block optimizes feature representation by shifting and capturing deep features through a combination of the Spatial Shift Block and depthwise separable convolutions. Dynamic adjustment of the loss weights at each layer approaches the optimal solution and enhances training stability. We validated LSU-Net on the UWMGI and MSD Colon datasets, and experimental results demonstrate that LSU-Net outperforms most state-of-the-art segmentation architectures. Copyright © 2025, The Authors. All rights reserved.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Multiple Instance Detection Networks With Adaptive Instance Refinement

引用

IEEE TRANSACTIONS ON MULTIMEDIA 2023年 25卷 267-279页

作者： Wu, Zhihao Wen, Jie Xu, Yong Yang, Jian Zhang, David Harbin Inst Technol Biocomp Res Ctr Shenzhen 518055 Peoples R China Shenzhen Key Lab Visual Object Detect & Recognit Shenzhen 518055 Peoples R China Pengcheng Lab Shenzhen 518055 Peoples R China Nanjing Univ Sci & Technol PCA Lab Key Lab Intelligent Percept & Syst High Dimens Inf Minist Educ Nanjing 210094 Peoples R China Nanjing Univ Sci & Technol Sch Comp Sci & Engn Jiangsu Key Lab Image & Video Understanding Social Nanjing 210094 Peoples R China Chinese Univ Hong Kong Shenzhen Sch Data Sci Shenzhen 518172 Peoples R China Shenzhen Res Inst Big Data Shenzhen 518172 Peoples R China Shenzhen Inst Artificial Intelligence & Robot Soc Shenzhen 518172 Peoples R China

Weakly supervised object detection (WSOD) aims to train object detectors by using only image-level annotations. Many recent works on WSOD adopt multiple instance detection networks (MIDN), which usually generate a certain number of proposals and regard proposal classification as a latent model learning within image classification. However, these methods tend to detect salient object, salient object parts and clustered objects due to lack of instance-level annotations during training. Thus a core issue is how to guarantee that the network learn as many objects with precise bounding boxes as possible. In this paper, we address this issue by exploiting the potential of proposal scores during training. We propose an adaptive instance refinement (AIR) framework with three novel designs, which can be integrated with MIDN into a single network. Specifically, adaptive instance mining attempts to discover all positive instances according to the score distribution of proposals and their spatial similarity. Adaptive score modulation dynamically adjusts proposal scores to make the network focus more on instances with different difficulties in different training iterations. Adaptive knowledge refinement distills important information from all previous stages by the weighted average of proposal scores. The experimental results on the PASCAL VOC 2007 and 2012 benchmarks and the MS COCO benchmark demonstrate that AIR significantly improves the performance of the original MIDN and achieves the state-of-the-art results.

关键词： Weakly supervised object detection multiple instance detection networks instance refinement framework proposal score

来源：评论

学校读者我要写书评

暂无评论

AUCSeg: AUC-oriented Pixel-level Long-tail Semantic Segmentation

arXiv

引用

arXiv 2024年

作者： Han, Boyu Xu, Qianqian Yang, Zhiyong Bao, Shilong Wen, Peisong Jiang, Yangbangyan Huang, Qingming Key Lab. of Intelligent Information Processing Institute of Computing Technology CAS China School of Computer Science and Tech. University of Chinese Academy of Sciences China Peng Cheng Laboratory China Key Laboratory of Big Data Mining and Knowledge Management CAS China

The Area Under the ROC Curve (AUC) is a well-known metric for evaluating instance-level long-tail learning problems. In the past two decades, many AUC optimization methods have been proposed to improve model performance under long-tail distributions. In this paper, we explore AUC optimization methods in the context of pixel-level long-tail semantic segmentation, a much more complicated scenario. This task introduces two major challenges for AUC optimization techniques. On one hand, AUC optimization in a pixel-level task involves complex coupling across loss terms, with structured inner-image and pairwise inter-image dependencies, complicating theoretical analysis. On the other hand, we find that mini-batch estimation of AUC loss in this case requires a larger batch size, resulting in an unaffordable space complexity. To address these issues, we develop a pixel-level AUC loss function and conduct a dependency-graph-based theoretical analysis of the algorithm’s generalization ability. Additionally, we design a Tail-Classes Memory Bank (T-Memory Bank) to manage the significant memory demand. Finally, comprehensive experiments across various benchmarks confirm the effectiveness of our proposed AUCSeg method. The code is available at https://***/boyuh/AUCSeg. © 2024, CC BY.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

PP-LDG: A Medical Privacy-Preserving Labeled data Generation Framework

PP-LDG: A Medical Privacy-Preserving Labeled Data Generation...

引用

Medical Artificial Intelligence (MedAI), IEEE International Conference on

作者： Haoxiang Yuan Xiaochen Yuan Xiuli Bi Weisheng Li Guoyin Wang Bin Xiao Chongqing Key Laboratory of Image Cognition School of Computer Science and Technology Chongqing University of Posts and Telecommunications Chongqing China Faculty of Applied Sciences Macao Polytechnic University Macao China Key Laboratory of Cyberspace Big Data Intelligent Security Ministry of Education Chongqing China

ISBN: (数字)9798350377613

ISBN: (纸本)9798350377620

The rapid development of deep learning has led to an increasing demand for data. However, such data are scarce in many fields and often contain private and sensitive information, such as medical images. The fact that data owners are unwilling to share these high-privacy data further exacerbates the scarcity of data. A prevalent research direction is to combine differential privacy and image generation, by which we can obtain a large amount of synthetic data with a similar distribution to the original dataset, and that synthetic dataset does not compromise the privacy of the original dataset. However, these methods cannot provide labels for the generated data, limiting the subsequent uses of the data. To solve the problem, we propose a generic privacy-preserving labeled data generation framework named PP-LDG. It is the first data publishing framework to generate labeled data while not disclosing the privacy of original data. We theoretically demonstrate that our proposed framework can provide strict privacy guarantees with differential privacy and demonstrate the effectiveness of the synthetic dataset obtained from the framework under practical privacy budgets through extensive experiments.

关键词： Deep learning Differential privacy Accuracy Limiting Image synthesis Publishing data collection Synthetic data Biomedical imaging Image classification

来源：评论

学校读者我要写书评

暂无评论

No-Reference Stereoscopic Image Quality Assessment Based on Binocular Collaboration

SSRN

引用

SSRN 2024年

作者： Guo, Wenzhong Wang, Hanling Ke, Xiao Fujian Provincial Key Laboratory of Networking Computing and Intelligent Information Processing College of Computer and Data Science Fuzhou University Fujian Fuzhou350116 China Engineering Research Center of Big Data Intelligence Ministry of Education Fuzhou University Fuzhou350116 China

Stereoscopic images usually consist of a left view, a right view, and depth information. Stereoscopic/3D image quality assessment (SIQA) is often more challenging compared to 2D-IQA due to the usual scene differences between left and right views and the complexity of the binocular vision fusion process. To address the problem of quality prediction bias of multi-distortion images, we investigated the visual physiology and the processing of visual information by the primary visual cortex of the human brain and proposed a no-reference stereoscopic image quality evaluation method. The method mainly includes an innovative end-to-end NR-SIQA neural network with a picture patch generation algorithm. The algorithm generates a saliency map by fusing the left and right views and then guides the image cropping in the database based on the saliency map. The proposed models are validated and compared based on publicly available databases. The results show that the model and algorithm together outperform the state-of-the-art NR-SIQA metric in the LIVE 3D database and the WIVC 3D database, and have excellent results in the specific noise metric. The model generalization experiments demonstrate a certain degree of generality of our proposed model. © 2024, The Authors. All rights reserved.

关键词： database systems

来源：评论

学校读者我要写书评

暂无评论

data-Driven Discovery of Partial Differential Equations Based on Temporal Relationships

SSRN

引用

SSRN 2024年

作者： Zhang, Xiaoxia Mao, Hao Guan, Junsheng Liu, Yanjun Wang, Guoyin Chongqing Key Laboratory of Computational Intelligence Chongqing University of Posts and Telecommunications Chongqing400065 China Key Laboratory of Big Data Intelligent Computing Chongqing University of Posts and Telecommunications Chongqing400065 China School of Mathematical Sciences Chongqing Normal University Chongqing401331 China

Partial differential equations (PDEs) are crucial for describing and understanding various physical phenomena. As observational datasets become increasingly rich, data-driven approaches for discovering PDEs have become mainstream. PDEs describe how variables in physical systems or processes evolve over time and space, and their temporal nature plays a crucial role in modeling many physical phenomena and engineering problems. They can provide important information for equation identification and discovery. In our approach, we introduce Long Short-Term Memory (LSTM) and self-attention mechanisms to extract temporal features embedded in the data. The LSTM layer is used to capture temporal information from input sequences, while the self-attention mechanism enhances or suppresses different feature representations by computing attention weights. Through iterative training of neural networks and leveraging automatic differentiation to obtain derivative terms, we construct a candidate function library. Finally, by employing sparse regression, we determine the coefficient vector to determine the form of the equation. We demonstrate the feasibility of our method through experiments on multiple datasets. © 2024, The Authors. All rights reserved.

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

Spatial Adaptive Filter Network With Scale-Sharing Convolution for Image Demoireing

引用

IEEE SIGNAL PROCESSING LETTERS 2024年 31卷 2495-2499页

作者： Xu, Yong Wei, Zhiyu Xu, Ruotao Zhou, Zihan Yu, Zhuliang South China Univ Technol Sch Comp Sci & Engn Guangzhou 510006 Peoples R China Guangdong Prov Key Lab Multimodal Big Data Intelli Guangzhou 519085 Peoples R China PaZhou Lab Guangzhou Guangzhou 510335 Peoples R China Inst Super Robot Huangpu Guangzhou 510555 Peoples R China South China Agr Univ Sch Coll Math & Informat Guangzhou Peoples R China South China Univ Technol Shien Ming Wu Sch Intelligent Engn Guangzhou 511442 Peoples R China South China Univ Technol Sch Automat Sci & Engn Guangzhou 510641 Peoples R China

Removing moire patterns is a challenging task as it is a spatially varying degradation that varies in shape, color and scale. Existing image restoration models often rely on static convolutional neural networks (CNNs)-based architectures, and hence potentially suboptimal for addressing the diverse manifestations of moire patterns across different images and spatial positions. To this end, we propose a spatially adaptive neural network for image demoireing. This network introduces a dual-branch filter prediction module engineered to predict pixel-wise adaptive filters that can process moire patterns of varying orientations and color-shift issues. To further tackle the challenge presented by scale variability, a scale-sharing convolution module is proposed, utilizing pixel-wise adaptive filters with multiple dilations to handle moire patterns of different sizes but similar shapes effectively. Upon extensive evaluations of three benchmark datasets, our model consistently outperforms existing methods, yielding a PSNR improvement of over 0.37dB across all evaluated datasets and providing additional benefits in terms of model size.

关键词： Convolution Feature extraction Adaptive filters Image color analysis Shape Tensors Semantics Moir & eacute pattern removal dynamic filtering spatial-varying processing

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：