检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Zhang, Boxuan Wang, Zengmao Du, Bo The School of Computer Science Wuhan University Wuhan430072 China The National Engineering Research Center for Multimedia Software School of Computer Science Institute of Artificial Intelligence The Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University Wuhan430072 China

The lack of object-level annotations poses a significant challenge for object detection in remote sensing images. To address this issue, active learning and semi-supervised learning techniques have been proposed to enhance the quality and quantity of annotations. Active learning focuses on selecting the most informative samples for annotation, while semi-supervised learning leverages the knowledge from unlabeled samples. In this paper, we propose a novel active learning method to boost semi-supervised object detection for remote sensing images with a teacher-student network, called SSOD-AT. The proposed method incorporates a RoI Comparison module (RoICM) to generate high-confidence pseudo-labels for Regions of Interest (RoIs). Meanwhile, the RoICM is utilized to identify the top-K uncertain images. To reduce redundancy in the top-K uncertain images for human labeling, a diversity criterion is introduced based on object-level prototypes of different categories using both labeled and pseudo-labeled images. Extensive experiments on DOTA and DIOR two popular datasets demonstrate that our proposed method outperforms state-of-the-art methods for object detection in remote sensing images. Compared with the best performance in the SOTA methods, the proposed method achieves 1% improvement at most cases in the whole active learning. © 2024, CC BY.

关键词： Remote sensing

来源：评论

学校读者我要写书评

暂无评论

Improving Acoustic Echo Cancellation by Mixing Speech Local and Global Features with Transformer

Improving Acoustic Echo Cancellation by Mixing Speech Local ...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Yajie Liu Xinmeng Xu Weiping Tu Yuhong Yang Li Xiao School of Computer Science National Engineering Research Center for Multimedia Software Wuhan University China Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University China

We propose MiT-Net, a novel mix-transformer neural network with a pyramid encoder operating in the time domain, for the task of acoustic echo cancellation. The MiT-Net formulates acoustic echo cancellation as a supervised speech separation problem, in which near-end speech is separated from a single microphone recording and sent to the far end, and consists of two key components. First, we apply a pyramid encoder, which adopts the coarse-to-fine structure, to extract the latent correlations between double-end signals and to fuse them in a multiscale manner. Second, we propose a mix-transformer, a combination of local and global attention in a parallel way, to leverage local and global speech information for separation. Experimental results show that the proposed method outperforms recent AEC methods in terms of objective evaluation metrics. In addition, exploring the correlation between speech local and global features by using the mix-transformer significantly improves the system performance and shows more robustness than the conventional transformer.

关键词： Echo cancellers Correlation Fuses System performance Transformers Feature extraction Acoustics

来源：评论

学校读者我要写书评

暂无评论

Bi-Causal: Group Activity Recognition via Bidirectional Causality

Bi-Causal: Group Activity Recognition via Bidirectional Caus...

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Youliang Zhang Wenxuan Liu Danni Xu Zhuo Zhou Zheng Wang National Engineering Research Center for Multimedia Software Institute of Artificial Intelligence School of Computer Science Wuhan University Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University of Technology National University of Singapore

ISBN: (数字)9798350353006

ISBN: (纸本)9798350353013

Current approaches in Group Activity Recognition (GAR) predominantly emphasize Human Relations (HRs) while often neglecting the impact of Human-Object Inter-actions (HOIs). This study prioritizes the consideration of both HRs and HOIs, emphasizing their interdependence. Notably, employing Granger Causality Tests reveals the presence of bidirectional causality between HRs and HOIs. Leveraging this insight, we propose a Bidirectional-Causal GAR network. This network establishes a causality commu-nication channel while modeling relations and interactions, enabling reciprocal enhancement between human-object interactions and human relations, ensuring their mutual consistency. Additionally, an Interaction Module is devised to effectively capture the dynamic nature of human-object interactions. Comprehensive experiments conducted on two publicly available datasets showcase the superiority of our proposed method over state-of-the-art approaches. Our project page: https://***/***/

关键词： computer vision Computational modeling Cause effect analysis Communication channels Activity recognition

来源：评论

学校读者我要写书评

暂无评论

Multi-speaker Direction of Arrival Estimation Using Audio and Visual Modalities with Convolutional Neural network

Multi-speaker Direction of Arrival Estimation Using Audio an...

引用

IEEE International Conference on Multimedia and Expo (ICME)

作者： Yulin Wu Ruimin Hu Xiaochen Wang National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University Wuhan China

In reality, audible and visible sound sources are closely aligned, and they can help humans locate sources exactly. To exploit the complementarity between audio and visual data in multi-speaker direction of arrival (DoA) estimation, we propose a novel network consisting of 3D convolution neural networks (3D-CNNs) and 2D-CNNs mixture networks with residual dense blocks. It has two main advantages: 1) both input audio and visual features are low-level signal representation: the real and imaginary parts of STFT coefficients for the audio feature and pixel coordinates for the visual feature, which can allow the network to learn to extract the most informative high-level features. 2) 3D-CNNs with the residual dense block are used for audio and visual feature mapping along the time and frequency axis. The following 2D-CNNs are to ensemble the high-level features along the DoA axis. Experimental results demonstrate promising SSL performance.

关键词：

来源：评论

学校读者我要写书评

暂无评论

HAPG-SAQAM: Human Auditory Perception Guided Spatial Audio Quality Assessment Metric

HAPG-SAQAM: Human Auditory Perception Guided Spatial Audio Q...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Yuanming Zheng Jiaxuan Yao Xiangyu Deng Yuhong Yang Ruiqi Liao Weiping Tu Cedar Lin National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University Wuhan China Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan China Guangdong OPPO Mobile Telecommunications Corp. China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Spatial audio quality evaluation is essential for applications like virtual and augmented reality, where accurate sound reproduction enhances user immersion. While subjective listening tests are the gold standard, they are costly and time-consuming. To address this, we propose HAPG-SAQAM, an objective metric for assessing timbre quality, spatial quality, and overall quality of binaural audio, guided by human auditory perception. Our contributions include: (1) the Multi-scale Auditory Guided Feature Extraction (MAGFE) module, incorporating gammatone frequency cepstral coefficients for better alignment with human perception; (2) Perceptual Weighted Loss (PWL), optimizing the weighting of timbre quality (TQ) and spatial quality (SQ) loss based on subjective test data; and (3) data augmentation techniques to enhance robustness by amplifying perceptual distortions. Experimental results show HAPG-SAQAM improves correlation with subjective scores by 10%, with ablation studies confirming the contributions of its components to enhanced spatial and overall audio quality.

关键词： Measurement Accuracy Spatial audio Performance gain Feature extraction Robustness Quality assessment Timbre Speech processing Optimization

来源：评论

学校读者我要写书评

暂无评论

Natural Human Skin‑Inspired Wearable and Breathable Nanofiber‑based Sensors with Excellent Thermal Management Functionality

引用

Advanced Fiber Materials 2024年第6期6卷 1955-1968页

作者： Peng Wang Xiaodan Li Guifen Sun Guoqing Wang Qing Han Chuizhou Meng Zhonghe Wei Yang Li School of Mechanical Engineering University of JinanJinan 250022China School of Information and Control Engineering China University of Mining and TechnologyXuzhou 221116China State Key Laboratory for Reliability and Intelligence of Electrical Equipment Engineering Research Center of Ministry of Education for Intelligent Rehabilitation Device and Detection TechnologyHebei Key Laboratory of Smart Sensing and Human‑Robot InteractionSchool of Mechanical EngineeringHebei University of TechnologyTianjin 300401China Shandong Provincial Key Laboratory of Network‑Based Intelligent Computing School of Information Science and EngineeringUniversity of JinanJinan 250022China Network Information Center Department of Computer ScienceQilu University of Technology(Shandong Academy of Sciences)Jinan 250353China School of Integrated Circuits Shandong UniversityJinan 250101China

Wearable sensors have been rapidly developed for application in various human monitoring ***,the wearing comfort and thermal properties of these devices have been largely ignored,and these characteristics urgently need to be ***,we develop a wearable and breathable nanofiber-based sensor with excellent thermal management functionality based on passive heat preservation and active Joule heating *** multifunctional device consists of a micropatterned carbon nanotube(CNT)/thermoplastic polyurethane(TPU)nanofiber electrode,a microporous ionic aerogel electrolyte and a microstructured Ag/TPU nanofiber *** to the presence of a supercapacitive sensing mechanism and the appli-cation of microstructuration,the sensor shows excellent sensing performance,with a sensitivity of 24.62 ***,due to the overall porous structure and hydrophobicity of TPU,the sensor shows good breathability(62 mm/s)and water repellency,with a water contact angle of 151.2°.In addition,effective passive heat preservation is achieved by combining CNTs with high solar absorption rates(85%)as the top layer facing the outside,aerogel with a low thermal conductivity(0.063 W m-1 k-1)as the middle layer for thermal insulation,and Ag with a high infrared reflectance rate as the bottom layer facing the *** warming,this material yields a higher temperature than ***,the active Joule heat-ing effect is realized by applying current through the bottom resistive electrode,which can quickly increase the temperature to supply controlled warming on *** proposed wearable and breathable sensor with tunable thermal properties is promising for monitoring and heat therapy applications in cold environments.

关键词： Wearable sensor Breathable Personal thermal management Electrospun nanofibers Joule heating

来源：评论

学校读者我要写书评

暂无评论

IQ-VFI: Implicit Quadratic Motion Estimation for Video Frame Interpolation

IQ-VFI: Implicit Quadratic Motion Estimation for Video Frame...

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Mengshun Hu Kui Jiang Zhihang Zhong Zheng Wang Yinqiang Zheng National Engineering Research Center for Multimedia Software Institute of Artificial Intelligence School of Computer Science Wuhan University Hubei Key Laboratory of Multimedia and Network Communication Engineering Harbin Institute of Technology Shanghai Artificial Intelligence Laboratory The University of Tokyo

ISBN: (数字)9798350353006

ISBN: (纸本)9798350353013

Advanced video frame interpolation (VFI) algorithms approximate intermediate motions between two input frames to synthesize intermediate frame. However, they struggle to handle complex scenarios with curvilinear motions since they overlook the latent acceleration information between the input frames. Moreover, the supervision of predicted motions is tricky because ground-truth motions are not available. To this end, we propose a novel frame-work for implicit quadratic video frame interpolation (IQ-VFI), which explores latent acceleration information and accurate intermediate motions via knowledge distillation. Specifically, the proposed IQ-VFI consists of an implicit acceleration estimation network (IANet) and a VFI back-bone, the former fully leverages spatio-temporal information to explore latent acceleration priors between two input frames, which is then used to progressively modulate linear motions from the latter into quadratic motions in coarse-to-fine manner. Furthermore, to encourage both components to distill more acceleration and motion cues oriented towards VFI, we propose a knowledge distillation strategy in which implicit acceleration distillation loss and implicit motion distillation loss are employed to adaptively guide latent acceleration priors and intermediate motions learning, respectively. Extensive experiments show that our proposed IQ-VFI can achieve state-of-the-art performances on various benchmark datasets.

关键词： Interpolation computer vision Accuracy Motion estimation Computational modeling Estimation Benchmark testing

来源：评论

学校读者我要写书评

暂无评论

Emotion Recognition Empowered Human-computer Interaction with Domain Adaptation network

引用

IEEE Transactions on Consumer Electronics 2024年

作者： Xu, Xu Fu, Chong Chen, Junxin Northeastern University School of Computer Science and Engineering Shenyang110004 China Dalian University of Technology China The School of Software Dalian116621 China Northeastern University School of Computer Science and Engineering Shenyang110004 China Northeastern University Key Laboratory of Intelligent Computing in Medical Image Ministry of Education Shenyang110819 China Ministry of Education Engineering Research Center of Security Technology of Complex Network System Shenyang110004 China Dalian University of Technology School of Software Dalian116621 China

Multi-modal emotion recognition plays a vital role in the human-computer interaction (HCI) for consumer electronics. Nowadays, many studies have developed multi-modal fusion algorithms for this ***, two challenging issues remain unsolved, i.e., inefficient multi-modal feature fusion and unclear distance in feature *** this end, we develop a novel framework, namely LAFDA-Net, for cross-subject emotion recognition using EEG and eye movement *** is based on low-rank fusion and domain adaptation *** specifically, the multi-modal signals are input into the feature extraction branch in parallel to generate ***, these features are fused by the low-rank fusion branch, reducing complexity and avoiding ***, the fused features are flattened and sent to the classification branch to determine the emotion *** training, these features are input into the domain adaptive branch to bridge the gap between the source domain and target *** benchmark datasets, i.e., SEED, SEED-IV, and SEED-V, are employed for performance *** results demonstrate that the proposed LAFDA-Net is robust, effective, and has advantages over peer methods. © 1975-2011 IEEE.

关键词： Emotion Recognition

来源：评论

学校读者我要写书评

暂无评论

MTGA: Multi-View Temporal Granularity Aligned Aggregation for Event-Based Lip-Reading

arXiv

引用

arXiv 2024年

作者： Zhang, Wenhao Wang, Jun Luo, Yong Yu, Lei Yu, Wei He, Zheng Shen, Jialie School of Computer Science National Engineering Research Center for Multimedia Software Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University China School of Electronics and Information Wuhan University China Department of Computer Science City University of London United Kingdom

Lip-reading is to utilize the visual information of the speaker’s lip movements to recognize words and sentences. Existing event-based lip-reading solutions integrate different frame rate branches to learn spatio-temporal features of varying granularities. However, aggregating events into event frames inevitably leads to the loss of fine-grained temporal information within frames. To remedy this drawback, we propose a novel framework termed Multi-view Temporal Granularity aligned Aggregation (MTGA). Specifically, we first present a novel event representation method, namely time-segmented voxel graph list, where the most significant local voxels are temporally connected into a graph list. Then we design a spatio-temporal fusion module based on temporal granularity alignment, where the global spatial features extracted from event frames, together with the local relative spatial and temporal features contained in voxel graph list are effectively aligned and integrated. Finally, we design a temporal aggregation module that incorporates positional encoding, which enables the capture of local absolute spatial and global temporal information. Experiments demonstrate that our method outperforms both the event-based and video-based lip-reading counterparts. Copyright © 2024, The Authors. All rights reserved.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

TextIM: Part-aware Interactive Motion Synthesis from Text

arXiv

引用

arXiv 2024年

作者： Fan, Siyuan Du, Bo Cai, Xiantao Peng, Bo Sun, Longling National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University China Institute of Artificial Intelligence Wuhan University China Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University China

In this work, we propose TextIM, a novel framework for synthesizing TEXT-driven human Interactive Motions, with a focus on the precise alignment of part-level semantics. Existing methods often overlook the critical roles of interactive body parts and fail to adequately capture and align part-level semantics, resulting in inaccuracies and even erroneous movement outcomes. To address these issues, TextIM utilizes a decoupled conditional diffusion framework to enhance the detailed alignment between interactive movements and corresponding semantic intents from textual descriptions. Our approach leverages large language models, functioning as a human brain, to identify interacting human body parts and to comprehend interaction semantics to generate complicated and subtle interactive motion. Guided by the refined movements of the interacting parts, TextIM further extends these movements into a coherent whole-body motion. We design a spatial coherence module to complement the entire body movements while maintaining consistency and harmony across body parts using a part graph convolutional network. For training and evaluation, we carefully selected and re-labeled interactive motions from HUMANML3D to develop a specialized dataset. Experimental results demonstrate that TextIM produces semantically accurate human interactive motions, significantly enhancing the realism and applicability of synthesized interactive motions in diverse scenarios, even including interactions with deformable and dynamically changing objects. Copyright © 2024, The Authors. All rights reserved.

关键词： Semantics

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：