检索结果-内蒙古大学图书馆

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Wei Li Hehe Fan Yongkang Wong Mohan Kankanhalli Yi Yang ReLER Lab CCAI and The State Key Laboratory of Brain-Machine Intelligence Zhejiang University China ReLER Lab CCAI Zhejiang University China School of Computing National University of Singapore Singapore

ISBN: (纸本)9798331314385

Recent advancements in image understanding have benefited from the extensive use of web image-text pairs. However, video understanding remains a challenge despite the availability of substantial web video-text data. This difficulty primarily arises from the inherent complexity of videos and the inefficient language supervision in recent web-collected video-text datasets. In this paper, we introduce Text-Only Pre-Alignment (TOPA), a novel approach to extend large language models (LLMs) for video understanding, without the need for pre-training on real video data. Specifically, we first employ an advanced LLM to automatically generate Textual Videos comprising continuous textual frames, along with corresponding annotations to simulate real video-text pairs. Then, these annotated textual videos are used to pre-align language-only LLMs with the video modality. To bridge the gap between textual and real videos, we employ the CLIP model as the feature extractor to align image and text modalities. During text-only pre-alignment, the continuous textual frames, encoded as a sequence of CLIP text features, are analogous to continuous CLIP image features, thus aligning the LLM with real video representation. Extensive experiments, including zero-shot evaluation and finetuning on various video understanding tasks, demonstrate that TOPA is an effective and efficient framework for aligning video modality with LLMs. In particular, without training on any video data, the TOPA-Llama2-13B model achieves a Top-1 accuracy of 51.0% on the challenging long-form video understanding benchmark, EgoSchema. This performance surpasses previous video-text pre-training approaches and is competitive with recent GPT-3.5-based video agents.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Chinese Expressive Long-dialogue Speech Dataset with Scripts

A Chinese Expressive Long-dialogue Speech Dataset with Scrip...

引用

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

作者： Jin Li Tianrui Wang Meng Ge Chenrui Cui Jiahui Zhao Jianrong Wang Longbiao Wang Jianwu Dang Tianjin Key Laboratory of Cognitive Computing and Application College of Intelligence and Computing Tianjin University Tianjin China College of Intelligence and Computing Tianjin University Tianjin China Huiyan Technology (Tianjin) Co. Ltd Tianjin China Shenzhen Institute of Advanced Technology Chinese Academy of Sciences Guangdong China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

With the advancement of large-scale models, the demand for emotionally rich, long-context, and highly natural communication in human-computer interaction increases. However, the exploration of long-context or script-level speech conversation tasks remains limited due to the lack of specific supervised data. To address this, we introduce a three-stage data processing pipeline for creating a Chinese expressive long-dialogue speech dataset with scripts (CELSDS). We collect videos from TV series, manually annotate speaker information for each character, apply Optical Character Recognition (OCR) to extract speech content, annotate episode summaries, and use a large language model (LLM) to generate sentence-level scenario descriptions. To our knowledge, this is the first Chinese long-context dialogue dataset that incorporates speaker and content annotations, script-level episode summaries, and sentence-level scenario details. Using this dataset, we develop a baseline model for both speech-to-script and script-to-speech generation tasks. The annotations and data production code are open-sourced at: https://***/lijin0120/CELSDS.

关键词： TV Annotations Pipelines Optical character recognition Speech recognition Production Signal processing Data processing Speech processing Videos

来源：评论

学校读者我要写书评

暂无评论

Region Attention Fine-tuning with CLIP for Few-shot Classification

Region Attention Fine-tuning with CLIP for Few-shot Classifi...

引用

IEEE International Conference on Multimedia and Expo (ICME)

作者： Guangxing Wu Junxi Chen Qiu Li Wentao Zhang Wei-Shi Zheng Ruixuan Wang School of Computer Science and Engineering Sun Yat-sen University Guangzhou China Key Laboratory of Machine Intelligence and Advanced Computing MOE Guangzhou China China United Network Communications Corporation Limited Guangdong Branch Guangzhou China Peng Cheng Laboratory Shenzhen China

ISBN: (数字)9798350390155

ISBN: (纸本)9798350390162

With the advancements in visual language models such as CLIP and their strong performance in zero-shot recognition, numerous CLIP-based methods have emerged in the field of few-shot classification. However, many of them do not fully leverage the abundant feature information within the CLIP visual encoder and overlook the issue of varying region-specific importance for image classification across different datasets. To address these limitations, we present an attention pooling-based framework for few-shot fine-tuning. Our framework enables the model to learn task-specific attention weights for image regions, while also incorporating background features and a consistency constraint to enhance training. As a result, our approach outperforms the state-of-the-art approaches on 11 benchmarks, demonstrating its effectiveness.

关键词： Training Visualization Benchmark testing Image classification

来源：评论

学校读者我要写书评

暂无评论

Class Incremental Learning with Task-Specific Batch Normalization and Out-of-Distribution Detection

arXiv

引用

arXiv 2024年

作者： Xie, Xuchen Qiu, Yiqiao Lin, Run Zheng, Weishi Wang, Ruixuan School of Computer Science and Engineering Sun Yat-sen University Guangzhou China Peng Cheng Laboratory Shenzhen China Key Laboratory of Machine Intelligence and Advanced Computing MOE Guangzhou China Department of Computer Science and Engineering University of California San Diego United States

This study focuses on incremental learning for image classification, exploring how to reduce catastrophic forgetting [1] of all learned knowledge when access to old data is restricted due to memory or privacy constraints. The challenge of incremental learning lies in achieving an optimal balance between plasticity, the ability to learn new knowledge, and stability, the ability to retain old knowledge. Based on whether the task identifier (task-ID) of an image can be obtained during the test stage, incremental learning for image classifcation is divided into two main paradigms, which are task incremental learning (TIL) and class incremental learning (CIL). The TIL paradigm has access to the task-ID, allowing it to use multiple task-specific classification heads selected based on the task-ID. Consequently, in CIL, where the task-ID is unavailable, TIL methods must predict the task-ID to extend their application to the CIL paradigm. Our previous method [2] for TIL adds task-specific batch normalization and classification heads incrementally. This work extends the method by predicting task-ID through an "unknown" class added to each classification head. The head with the lowest "unknown" probability is selected, enabling task-ID prediction and making the method applicable to CIL. The task-specific batch normalization (BN) modules effectively adjust the distribution of output feature maps across different tasks, enhancing the model’s plasticity. Moreover, since BN has much fewer parameters compared to convolutional kernels, most of the feature extractor’s parameters remain unchanged. By only modifying the BN layers as new tasks arrive, the model can effectively manage parameter growth while ensuring stability across tasks. In summary, the proposed method can efficiently control parameter growth while ensuring both the model’s plasticity and stability. The innovation of this study lies in the first-time introduction of task-specific BN into CIL and verifying the feasibili

关键词： Image classification

来源：评论

学校读者我要写书评

暂无评论

Boosting Lightweight Camouflaged Object Detection with Multi-Scale Context and Boundary Awareness

Boosting Lightweight Camouflaged Object Detection with Multi...

引用

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

作者： Zihan Xu Zheng Wang Haoyu Wang Cheng Liu Yan Zhou Meijun Sun College of Intelligence and Computing Tianjin University Tianjin China Tianjin Key Laboratory of Machine Learning Tianjin University Tianjin China School of Disaster and Emergency Medicine Tianjin University Tianjin China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

To adapt to the resource-limited environment, this study introduces the lightweight boundary-aware camouflaged object detection(COD) network LMABnet. We enhance the feature representation capability of the lightweight network through a multi-scale feature fusion architecture, while effectively avoids the model inflation and parameter redundancy of traditional COD methods. The LMABnet adopts the streamlined two-branch architecture that incorporates with the lightweight feature extraction module, which provides the multi-scale feature perception capability and reduces the loss of spatial information caused by rapid downsampling. In addition, based on the principle of "easy before difficult", the lightweight multi-scale feature fusion module discovers the key parts of the camouflaged object through the details. Then enhances the recognition of the object edges by using the edge-attention fusion module, so as to improve the accuracy performance of detection. Empirical evaluations conducted on the COD dataset demonstrate that the proposed methodology not only attains state-of-the-art performance but also accomplishes a significant reduction of 40% in parameter count relative to existing approaches.

关键词： Accuracy Image edge detection Scalability Redundancy Computer architecture Object detection Signal processing Feature extraction Real-time systems Speech processing

来源：评论

学校读者我要写书评

暂无评论

View-decoupled Transformer for Person Re-identification under Aerial-ground Camera Network

arXiv

引用

arXiv 2024年

作者： Zhang, Quan Wang, Lei Patel, Vishal M. Xie, Xiaohua Lai, Jianhuang School of Computer Science and Engineering Sun Yat-Sen University China Guangdong China Guangdong Province Key Laboratory of Information Security Technology Guangzhou China Key Laboratory of Machine Intelligence and Advanced Computing Ministry of Education China Department of Electrical and Computer Engineering Johns Hopkins University United States

Existing person re-identification methods have achieved remarkable advances in appearance-based identity association across homogeneous cameras, such as ground-ground matching. However, as a more practical scenario, aerial-ground person re-identification (AGPReID) among heterogeneous cameras has received minimal attention. To alleviate the disruption of discriminative identity representation by dramatic view discrepancy as the most significant challenge in AGPReID, the view-decoupled transformer (VDT) is proposed as a simple yet effective framework. Two major components are designed in VDT to decouple view-related and view-unrelated features, namely hierarchical subtractive separation and orthogonal loss, where the former separates these two features inside the VDT, and the latter constrains these two to be independent. In addition, we contribute a large-scale AGPReID dataset called CARGO, consisting of five/eight aerial/ground cameras, 5,000 identities, and 108,563 images. Experiments on two datasets show that VDT is a feasible and effective solution for AGPReID, surpassing the previous method on mAP/Rank1 by up to 5.0%/2.7% on CARGO and 3.7%/5.2% on AG-ReID, keeping the same magnitude of computational complexity. Our project is available at https://***/LinlyAC/VDT-AGPReID. © 2024, CC BY.

关键词： Antennas

来源：评论

学校读者我要写书评

暂无评论

Joint Edge and Regional Depth Enhancement Network for Camouflaged Object Detection

Joint Edge and Regional Depth Enhancement Network for Camouf...

引用

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

作者： Miao Qi Zheng Wang Cheng Liu Yan Zhou Meijun Sun College of Intelligence and Computing Tianjin University Tianjin China Tianjin Key Laboratory of Machine Learning Tianjin University Tianjin China School of Disaster and Emergency Medicine Tianjin University Tianjin China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Camouflaged object detection (COD) is a task of identifying and locating target objects that are camouflaged, masked, or confused. Research claims that depth cues can provide effective object location cues. However, depth images often contain noise interference, which may negatively affect object recognition. In addition, depth images also lack edge details, and most of previous works pay more attention to the integrity of the region, rather than the quality of the edge. To solve these two problems, we propose a joint edge and regional depth enhancement network (ERDENet) for Camouflaged Object Detection. The network first introduces the Locate and Generate Depth (LGD) module to locate the target region and generate its depth image. After that, the Feature Interactive Fusion (FIF) module is carried out, complementing the enhanced depth feature with the feature extracted from the original image, and then fuse the edge clues into it. Finally, we design a Muiti-modal Refinement Extraction (MRE) module to refine the feature to improve the detection performance. Extensive experiments show that our method has advantages and effectiveness.

关键词： Fuses Image edge detection Noise Object detection Interference Feature extraction Acoustics Object recognition Speech processing

来源：评论

学校读者我要写书评

暂无评论

Multi-scale Re-weighted Attention Feature Fusion for Non-Intrusive Load Monitoring

Multi-scale Re-weighted Attention Feature Fusion for Non-Int...

引用

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

作者： Lingxi Yang Meijun Sun Haowei Ran Yipu Liu Yan Zhou Zheng Wang College of Intelligence and Computing Tianjin University Tianjin China Tianjin Key Laboratory of Machine Learning Tianjin University Tianjin China School of Disaster and Emergency Medicine Tianjin University Tianjin China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Non-Intrusive Load Monitoring (NILM) addresses the challenge of disaggregating total energy consumption into individual appliance usage, which is essential for enhancing energy efficiency and managing smart grids. Existing methods often overlook the impact of window sizes on the separation of appliance-level power signals, leading to appliance power aliasing (APA), where a single total power curve corresponds to multiple operating conditions and appliance couplings. A larger window captures global features noise, but it also exacerbates APA, thereby enlarging the solution space. In contrast, a smaller window mitigates APA but it introduces noise and increases complexity. To address these limitations, we propose the Multi-Scale Re-weighted Attention Feature Fusion (MRAFF) model, which employs a multi-scale framework to manage temporal variations and incorporates a re-weighted attention module to enhance multi-scale interactions, thereby improving feature representation. Experimental results demonstrate that our MRAFF model significantly outperforms baseline models on the REDD and UK-DALE datasets, achieving improvements ranging from 2.27% to 10.5% in MAE and from 33.74% to 58.34% in SAE.

关键词： Couplings Load monitoring Energy consumption Noise Energy efficiency Distance measurement Smart grids Complexity theory Speech processing Load modeling

来源：评论

学校读者我要写书评

暂无评论

High-resolution and large-size stretchable electrodes based on patterned silver nanowires composites

引用

Nano Research 2022年第5期15卷 4590-4598页

作者： Yong Lin Qingsong Li Chen Ding Jiayi Wang Wei Yuan Zhiyuan Liu Wenming Su Zheng Cui Printable Electronics Research Centre Suzhou Institute of Nano-Tech and Nano-BionicsChinese Academy of SciencesSuzhou 215123China CAS Key Laboratory of Human-Machine Intelligence-Synergy Systems Shenzhen Institute of Advanced TechnologyChinese Academy of Sciences(CAS)Shenzhen 518055China School of Nano-Tech and Nano-Bionics University of Science and Technology of ChinaHefei 230026China

Electrodes based on composites of silver nanowires(AgNWs)and elastic polymers have been widely studied and applied in various stretchable electronic ***,due to the high aspect ratio of nanowires,the patterning of AgNW-based composite electrodes remains a huge challenge,especially for high-resolution complex circuit wiring on large-size elastic *** this paper,we propose a method for preparing large-size stretchable circuit boards with high-resolution electrodes by the combination of screen printing and vacuum filtration of AgNWs/polydimethylsiloxane(PDMS)*** as-prepared stretchable electrodes have smooth edges with patterning resolution up to-50 μ*** conductivity of the composite electrode can be precisely controlled by varying deposition densities of AgNWs and have reached to 1.07 x 10^(4) S/cm when the deposition density was 2.0 mg/cm^(2).In addition,the uniformity of conductivity and the resistance-strain characteristics of composite electrodes were systematically evaluated with different AgNWs deposition *** composite electrodes have been successfully employed to construct a large-size programmable display system and an 18-channel surface electromyography(EMG)recording,showing great potentials for some strain-insensitive stretchable circuits in wearable and health-related electronic applications.

关键词： AgNWs patterning silver nanowires AgNWs/PDMS composite stretchable circuit board high-density electrode array

来源：评论

学校读者我要写书评

暂无评论

CoLRP: A Contrastive Learning Abstractive Text Summarization Method with ROUGE Penalty

CoLRP: A Contrastive Learning Abstractive Text Summarization...

引用

International Joint Conference on Neural Networks (IJCNN)

作者： Caidong Tan Xiao Sun AnHui Province Key Laboratory of Affective Computing and Advanced Intelligent Machine School of computer and information Hefei University of Technology Hefei China

Contrastive learning can reduce the impact of ex-posure bias associated with training using maximum likelihood estimation, which aims to pull together positive samples to increase the likelihood of high-quality summaries and push away irrelevant negative samples to reduce the likelihood of low-quality summaries. In contrastive learning-based text summarization methods, a standard method for selecting positive and negative samples is randomly selected within a batch. This method can lead to sampling bias to the extent that the consistency of the representation space is compromised. Therefore, we propose a new method to penalize false negatives based on ROUGE metric scores as weights to sample from the dynamic output of the model training process. The method calculates ROUGE metric scores for penalizing false negatives in real-time and can distinguish between positive and negative samples to ensure spatial consistency and alleviate exposure bias. Experimental results on XSum, CNN/DM, and Multi-News datasets show that our approach effectively improves the performance of the latest text summarization pre-training models.

关键词：

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：