检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Liu, Xingxian Xu, Yajing Pattern Recognition & Intelligent System Laboratory Beijing University of Posts and Telecommunications Beijing China

Query-focused meeting summarization(QFMS) aims to generate a specific summary for the given query according to the meeting transcripts. Due to the conflict between long meetings and limited input size, previous works mainly adopt extract-then-summarize methods, which use extractors to simulate binary labels or ROUGE scores to extract utterances related to the query and then generate a summary. However, the previous approach fails to fully use the comparison between utterances. To the extractor, comparison orders are more important than specific scores. In this paper, we propose a Ranker-Generator framework. It learns to rank the utterances by comparing them in pairs and learning from the global orders, then uses top utterances as the generator’s input. We show that learning to rank utterances helps to select utterances related to the query effectively, and the summarizer can benefit from it. Experimental results on QMSum show that the proposed model outperforms all existing multi-stage models with fewer parameters. © 2023, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Foreground Prediction for Image Composition with Local and Global Feature Fusion 16

Foreground Prediction for Image Composition with Local and G...

引用

2024 16th International Conference on Graphics and Image Processing, ICGIP 2024

作者： Sun, Liliang He, Yuanlie Li, Wensheng Feng, Fujian Liang, Yihui School of Computer Guangdong University of Technology Guangzhou510000 China School of Computer Science Zhongshan Institute University of Electronic Science and Technology of China Zhongshan528400 China Guizhou Key Laboratory of Pattern Recognition and Intelligent System Guizhou Minzu University Guiyang550025 China

ISBN: (数字)9781510688780

ISBN: (纸本)9781510688773

This paper focuses on the image composition of transparent objects, where existing image matting methods suffer from composition errors due to the lack of accurate foreground during the composition process. We propose a foreground prediction model named ALGM, which leverages the local feature extraction capabilities of Convolutional Neural Networks (CNNs) and incorporates an attention mechanism for global information modeling. The proposed alpha-assisted foreground prediction module extracts foreground information from the original image and conveys it. The extracted foreground color information is combined with the deep structural features of the encoder and used for foreground color prediction. ALGM reduces image composition errors in the quantitative data from the Composition-1k dataset and improves the visual quality of composed images on the AIM-500 and Transparent-460 datasets. © 2025 SPIE.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Task Adaptive Parameter Fine-Tuning Based on Contribution Measure for Transfer Learning

SSRN

引用

SSRN 2023年

作者： Feng, Le Feng, Fujian Yang, Yuan Tan, Mian Wang, Lin Guizhou Key Laboratory of Pattern Recognition and Intelligent System Guizhou Minzu University Guiyang550025 China

Fine-tuning is an important transfer learning technique that has achieved significant success in various tasks lacking training data, and requires only a small number of training epochs to achieve satisfactory results. However, with the increasing complexity of the model scale and structure, designing appropriate fine-tuning schemes for specific target tasks becomes increasingly difficult. In this paper, a contribution measure criterion is used to quantify the importance of the pre-trained model parameters to the target task, providing a basis for selecting fine-tuning parameters. In addition, we found that the fine-tuning ratio vary depending on the specific target task. Therefore, we proposed an adaptive fine-tuning ratio search strategy to search the appropriate fine-tuning ratio for the given target task. Based on the above strategy, we propose an adaptive fine-tuning algorithm based on parameter contribution to customize the fine-tuning scheme for the target task. The experimental results show that the proposed algorithm can effectively quantify the contribution of model parameters, and our algorithm can adaptively adjust the fine-tuning ratio for the target task. Furthermore, our algorithm achieved state-of-the-art performance on seven publicly available visual classification datasets widely used in transfer learning. © 2023, The Authors. All rights reserved.

关键词： Parameter estimation

来源：评论

学校读者我要写书评

暂无评论

Self-Enhanced Training Framework for Referring Expression Grounding

Self-Enhanced Training Framework for Referring Expression Gr...

引用

IEEE International Conference on Image Processing

作者： Yitao Chen Ruoyi Du Kongming Liang Zhanyu Ma Pattern Recognition and Intelligent System Laboratory School of Artificial Intelligence Beijing University of Posts and Telecommunications Beijing

Weakly-supervised referring expression grounding (REG) aims at locating the image region described by a query sentence, where the mapping between the referential region and query is not available during the training stage. Noticing the significant gap between the fully- and weakly-supervised approaches, we develop a Self-Enhanced Training(SET) framework in this paper. Specifically, we first train the network under a weakly-supervised setting. Then, the model outputs are collected and filtered according to the confidence score and serve as pseudo-labels. Finally, with the help of these pseudo-labels, we tune the model under a fully-supervised setting. The SET framework provides a simple way of generating pseudo-labels that build a bridge between weak and full supervision. Experimental results demonstrate that model trained through our SET framework outperforms existing traditional methods on RefCOCO, RefCOCO+, and RefCOCOg datasets. The code is available at https://***/HTDL98/SET-framework.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Relation-Aware Learning for Multi-Task Multi-Agent Cooperative Games

引用

IEEE Transactions on Games 2024年 1-12页

作者： Yu, Yang Yang, Likun Guo, Zhourui Ren, Yongjian Yin, Qiyue Zhang, Junge Huang, Kaiqi Center for Research on Intelligent System and Engineering Institute of Automation Chinese Academy of Sciences Beijing China Center for Research on Intelligent System and Engineering and National Laboratory of Pattern Recognition Institute of Automation Chinese Academy of Sciences Beijing China

Collaboration among multiple tasks is advantageous for enhancing learning efficiency in multi-agent reinforcement learning. To guide agents in cooperating with different teammates in multiple tasks, contemporary approaches encourage agents to exploit common cooperative patterns or identify the learning priorities of multiple tasks. Despite the progress made by these methods, they all assume that all cooperative tasks to be learned are related and desire similar agent policies. This is rarely the case in multi-agent cooperation, where minor changes in team composition can lead to significant variations in cooperation, resulting in distinct cooperative strategies compete for limited learning resources. In this paper, to tackle the challenge posed by multi-task learning in potentially competing cooperative tasks, we propose a novel framework called Relation-Aware Learning (RAL). RAL incorporates a relation awareness module in both task representation and task optimization, aiding in reasoning about task relationships and mitigating negative transfers among dissimilar tasks. To assess the performance of RAL, we conduct a comparative analysis with baseline methods in a multi-task StarCraft environment. The results demonstrate the superiority of RAL in multi-task cooperative scenarios, particularly in scenarios involving multiple conflicting tasks. Index Terms—Cooperation games, multi-task learning, reinforcement learning. IEEE

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Trimap generation with background for natural image matting 3

Trimap generation with background for natural image matting

引用

3rd International Conference on Optics and Machine Vision, ICOMV 2024

作者： Fu, Qian Liang, Yihui Kun, Zou Feng, Fujian Xu, Xiang School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China School of Computer Science Zhongshan Institute University of Electronic Science and Technology of China Zhongshan China Guizou Key Laboratory of Pattern Recognition and Intelligent System Guizhou Minzu University Guiyang China

ISBN: (纸本)9781510680319

Image matting is a widely-used image processing technique that aims at accurately separating foreground from an image. However, this is a challenging and ill-posed problem that demands additional input, such as trimaps and background images, for providing prior knowledge. However, the manual annotation of trimaps require lots of labor, limiting the application of trimap-based methods. Some trimap-free methods explore alternatives with low labor requirements by utilizing captured background images, including background-based methods. However, the quality of alpha mattes predicted by trimap-free methods still fall short of trimap-based methods. To reduce the performance gap between background-based and trimap-based methodes, we present Trimap Generation from Background Image (TG-BG) method which can generate trimaps from the input image and a captured background image. It provides an economical solution to facilitate the application of trimap-based methods, allowing for low-cost and high-quality alpha matte predictions. TP-BG leverages a ViT backbone for feature extraction and employs the Image and Background Detail Fusion Stream (IBDFS) to capture multi-scale detail information. The introduction of foreground impact loss encourages the network to pay more attention to the foreground in the image. We validate the trimap prediction performance of TP-BG by comparing the alpha matte quality obtained by background-based methods and that obtained by trimap-based methods integrated with TP-BG. The experimental results demonstrate that TP-BG can generate high-quality trimap from a background image, and trimap-based methods integrated with TP-BG outperform the state-of-the-art background-based methods in terms of four alpha matte quality metrics. © 2024 SPIE.

关键词： Costs

来源：评论

学校读者我要写书评

暂无评论

Video matting based on local-global features fusion 4

Video matting based on local-global features fusion

引用

4th International Conference on Machine Learning and Computer Application, ICMLCA 2023

作者： Dong, Niuniu Liang, Yihui Zou, Kun Li, Wensheng Feng, Fujian School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China School of Computer Science Zhongshan Institute University of Electronic Science and Technology of China Zhongshan China Guizhou Key Laboratory of Pattern Recognition and Intelligent System Guizhou Minzu University Guiyang China

ISBN: (数字)9781510680265

ISBN: (纸本)9781510680258

Video matting aims at accurately separating foreground from videos. Recent video matting researches pursue to eliminate auxiliary inputs. However, due to the limited ability of extracting global correlation features, these methods suffer from performance degradation when dealing with complex scenes or natural background videos. To address this challenge, we propose a video matting method called Video Matting Based on Local-Global Features Fusion (VMBLGFF) which can extract both comprehensive global correlation features and local subtle features. VMBLGFF contains two closely connected networks: a transformer network that utilizes window and global attention mechanisms to obtain global correlation features within and cross windows, and a fusion network that integrates local subtle features into the global correlation features to supplement the local detail information which may be overlooked by the attention mechanisms. VMBLGFF alleviates the issue of limiting global correlation features and has been benchmarked on both synthetic and real datasets, and the results demonstrate that VMBLGFF improves the quality of video matting and exhibits good generalization performance. © 2024 SPIE.

关键词： Computer vision

来源：评论

学校读者我要写书评

暂无评论

Adaptive Face recognition Using Adversarial Information Network

arXiv

引用

arXiv 2023年

作者： Wang, Mei Deng, Weihong Pattern Recognition and Intelligent System Laboratory School of Artificial Intelligence Beijing University of Posts and Telecommunications Beijing100876 China

In many real-world applications, face recognition models often degenerate when training data (referred to as source domain) are different from testing data (referred to as target domain). To alleviate this mismatch caused by some factors like pose and skin tone, the utilization of pseudo-labels generated by clustering algorithms is an effective way in unsupervised domain adaptation. However, they always miss some hard positive samples. Supervision on pseudo-labeled samples attracts them towards their prototypes and would cause an intra-domain gap between pseudo-labeled samples and the remaining unlabeled samples within target domain, which results in the lack of discrimination in face recognition. In this paper, considering the particularity of face recognition, we propose a novel adversarial information network (AIN) to address it. First, a novel adversarial mutual information (MI) loss is proposed to alternately minimize MI with respect to the target classifier and maximize MI with respect to the feature extractor. By this min-max manner, the positions of target prototypes are adaptively modified which makes unlabeled images clustered more easily such that intra-domain gap can be mitigated. Second, to assist adversarial MI loss, we utilize a graph convolution network to predict linkage likelihoods between target data and generate pseudo-labels. It leverages valuable information in the context of nodes and can achieve more reliable results. The proposed method is evaluated under two scenarios, i.e., domain adaptation across poses and image conditions, and domain adaptation across faces with different skin tones. Extensive experiments show that AIN successfully improves cross-domain generalization and offers a new state-of-the-art on RFW dataset. © 2023, CC BY.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

Geometric Numerical Integral Method in Compact Lie Group 6

Geometric Numerical Integral Method in Compact Lie Group

引用

6th IEEE International Conference on intelligent Computing and Signal Processing, ICSP 2021

作者： Liu, Chao Yang, Shengyi Guizhou Minzu University Key Laboratory of Pattern Recognition Intelligent System of Guizhou Province Gui Yang China

ISBN: (纸本)9780738143705

Three dimensions special orthogonal group SO (3) is widely used to describe the rotation kinematics of the rigid body without local coordinates, which can avoid rotation singularity and unwinding in traditional methods. Propagating the rotation kinematics in SO (3) with a specific geometric integration method is not only to obtained numerical results with improved qualitative behavior, but also provided more accurate long-time integration results. While many studies have focused on geometric integration algorithms to preserve the geometric structure, this work has the additional objective of studying result accuracy. Integral curves on SO (3) obtained using the third-order Crouch-Grossman Lie group method are compared with numerical results using the third-order RKMK algorithms, the exponential coordinates method, and the simple projection method. Results show that the use of the Crouch-Grossman Lie group method better preserves the geometric structure of SO (3) for the larger time steps considered. It is also found that the third-order Crouch-Grossman algorithm is more accurate than the RKMK except for the smallest time step used. © 2021 IEEE.

关键词： Geometry

来源：评论

学校读者我要写书评

暂无评论

Research on Teaching Reform of 'Motor and Drive'Based on Matlab Simulation 2

Research on Teaching Reform of 'Motor and Drive'Based on Mat...

引用

2nd International Conference on Education, Knowledge and Information Management, ICEKIM 2021

作者： Chengwei, Zhang Yihua, Liang Liu, Chao Yang, Shengyi Key Laboratory of Pattern Recognition and Intelligent System of Guizhou Province Guizhou Minzu University Gui Yang China

ISBN: (纸本)9781728168340

'Motor and Drive' is a professional basic course with strong theory and wide application in practical engineering. Its theory is abstract and formula is complex, which is considered by teachers and automation students as one of the professional courses difficult to teach and learn. In order to better teaching and learning, teaching reform is provided. The teaching content is optimized, the teaching methods, experimental methods and assessment methods are reformed based on matlab simulation. It is helpful to improve the teaching quality, stimulate students' learning enthusiasm and solve practical problems. The teaching reform foundation of 'motor and drive' is provided in this paper. © 2021 IEEE.

关键词： Teaching

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：