检索结果-内蒙古大学图书馆

ieee International symposium on Circuits and Systems (ISCAS)

作者： Liu, Ye Hao, Shuang Huang, Kun Yang, Minghui Huang, Zili Qi, Xiuyuan Li, Yiting Zhou, Liang Long, Yu Zhou, Jun Univ Elect Sci & Technol China Chengdu Peoples R China

ISBN: (纸本)9798350330991;9798350331004

Optical flow plays an extremely important role in the field of computer vision and extremely high real-time performance is required especially in autonomous driving. Traditional optical flow methods generally improve the accuracy of optical flow through the image pyramid technique. However, the introduction of the image pyramid increases the computational complexity. Additionally, the data relationships between pyramid layers result in strong data dependencies, making it difficult to accelerate using parallel processing, and it is challenging to meet real-time requirements in practical scenarios. To address this issue, in this paper, we propose an FPGA-based ultra-high performance and scalable optical flow hardware accelerator with several techniques, including an adaptive optical flow computation technique based on dynamic direction prediction to reduce computation without accuracy degradation, a highly scalable computing architecture with configurable numbers of PEs to improve the flexibility and hardware utilization under different hardware resource constraints, and a reconfigurable pyramid-layer pipeline technique to improve performance and reduce memory size. The proposed hardware accelerator was implemented and evaluated on a Xilinx FPGA ZCU104 achieving ultra-high performance (405 FPS) while maintaining high accuracy (AEE 0.64).

关键词： Optical Flow Hardware Accelerator image Pyramid FPGA

来源：评论

学校读者我要写书评

暂无评论

Edge-Aware Autoencoder Design for Real-Time Mixture-of-Experts image Compression 56

Edge-Aware Autoencoder Design for Real-Time Mixture-of-Exper...

引用

56th ieee International symposium on Circuits and Systems (ISCAS)

作者： Fleig, Elvira Geistert, Jonas Bochinski, Erik Jongebloed, Rolf Sikora, Thomas Tech Univ Berlin Commun Syst Grp Berlin Germany

ISBN: (纸本)9781665451093

Steered-Mixtures-of-Experts (SMoE) models provide sparse, edge-aware representations, applicable to many use-cases in image processing. This includes denoising, super-resolution and compression of 2D- and higher dimensional pixel data. Recent works for image compression indicate that compression of images based on SMoE models can provide competitive performance to the state-of-the-art. Unfortunately, the iterative model-building process at the encoder comes with excessive computational demands. In this paper we introduce a novel edge-aware Autoencoder (AE) strategy designed to avoid the time-consuming iterative optimization of SMoE models. This is done by directly mapping pixel blocks to model parameters for compression, in spirit similar to recent works on "unfolding" of algorithms, while maintaining full compatibility to the established SMoE framework. With our plug-in AE encoder, we achieve a quantum-leap in performance with encoder run-time savings by a factor of 500 to 1000 with even improved image reconstruction quality. For image compression the plug-in AE encoder has real-time properties and improves RD-performance compared to our previous works.

关键词： image Compression image processing Autoencoder Sparse Representation Steered Mixture-of-Experts

来源：评论

学校读者我要写书评

暂无评论

A Lightweight Convolutional Dual-Channel Transformer Framework for Efficient Part Segmentation of Point Cloud Data 4

A Lightweight Convolutional Dual-Channel Transformer Framewo...

引用

4th International symposium on Artificial intelligence and Intelligent Manufacturing, AIIM 2024

作者： Li, Jianhua Xu, Yuanping Zhang, Chaolong Kong, Chao Jin, Jin Wang, Weiye Xu, Zhijie Guo, Benjun Tang, Dan School of Software Engineering Chengdu University of Information Technology Chengdu China School of Advanced Technology Xi'an Jiaotong-Liverpool University Suzhou China

ISBN: (纸本)9798331541729

In recent years, transformer-based models have made significant breakthroughs in natural language processing and computer vision. However, these models have encountered challenges when dealing with point cloud data because the irregular and disordered structure of point cloud data leads to a huge computational and memory burden. To address this problem, this paper proposes a Multidimensional Convolution-Dual Channel Transformer Network for efficient processing of point cloud data. The MCDTN framework consists of two branches: the main channel enhances the modeling of cross-channel features through dynamic attention to optimize feature representation;the auxiliary channel further improves fine-grained segmentation capabilities through encoders, multi-scale information interaction, and spatial attention. Experimental results show that MCDTN performs excellently in shape classification and part segmentation tasks, effectively reducing computational costs. © 2024 ieee.

关键词： Natural language processing systems

来源：评论

学校读者我要写书评

暂无评论

Raformer: Redundancy-Aware Transformer for Video Wire Inpainting

引用

ieee TRANSACTIONS ON image processing 2025年 34卷 1795-1809页

作者： Ji, Zhong Su, Yimu Zhang, Yan Hou, Jiacheng Pang, Yanwei Han, Jungong Tianjin Univ Sch Elect & Informat Engn Tianjin Key Lab Brain Inspired Intelligence Techno Tianjin 300072 Peoples R China Shanghai Artificial Intelligence Lab Shanghai 200232 Peoples R China Tsinghua Univ Dept Automat Beijing 100084 Peoples R China

Video Wire Inpainting (VWI) is a prominent application in video inpainting, aimed at flawlessly removing wires in films or TV series, offering significant time and labor savings compared to manual frame-by-frame removal. However, wire removal poses greater challenges due to the wires being longer and slimmer than objects typically targeted in general video inpainting tasks, and often intersecting with people and background objects irregularly, which adds complexity to the inpainting process. Recognizing the limitations posed by existing video wire datasets, which are characterized by their small size, poor quality, and limited variety of scenes, we introduce a new VWI dataset with a novel mask generation strategy, namely Wire Removal Video Dataset 2 (WRV2) and Pseudo Wire-Shaped (PWS) Masks. WRV2 dataset comprises over 4,000 videos with an average length of 80 frames, designed to facilitate the development and efficacy of inpainting models. Building upon this, our research proposes the Redundancy-Aware Transformer (Raformer) method that addresses the unique challenges of wire removal in video inpainting. Unlike conventional approaches that indiscriminately process all frame patches, Raformer employs a novel strategy to selectively bypass redundant parts, such as static background segments devoid of valuable information for inpainting. At the core of Raformer is the Redundancy-Aware Attention (RAA) module, which isolates and accentuates essential content through a coarse-grained, window-based attention mechanism. This is complemented by a Soft Feature Alignment (SFA) module, which refines these features and achieves end-to-end feature alignment. Extensive experiments on both the traditional video inpainting datasets and our proposed WRV2 dataset demonstrate that Raformer outperforms other state-of-the-art methods. Our codes and the WRV2 dataset will be made available at: https://***/Suyimu/WRV2.

关键词： Wire Transformers Films Visualization Training TV Production Manuals Context modeling computational modeling Video wire inpainting video inpainting video transformer redundant elimination

来源：评论

学校读者我要写书评

暂无评论

ADStereo: Efficient Stereo Matching With Adaptive Downsampling and Disparity Alignment

引用

ieee TRANSACTIONS ON image processing 2025年 34卷 1204-1218页

作者： Wang, Yun Li, Kunhong Wang, Longguang Hu, Junjie Wu, Dapeng Oliver Guo, Yulan City Univ Hong Kong Dept Comp Sci Kowloon Peoples R China Sun Yat Sen Univ Sch Elect & Commun Engn Shenzhen Campus Shenzhen 518107 Peoples R China Aviat Univ Air Force Coll Elect Sci & Technol Changchun 130012 Peoples R China Shenzhen Inst Artificial Intelligence & Robot Soc Shenzhen 518129 Peoples R China

The balance between accuracy and computational efficiency is crucial for the applications of deep learning- based stereo matching algorithms in real-world scenarios. Since matching cost aggregation is usually the most computationally expensive component, a common practice is to construct cost volumes at a low resolution for aggregation and then directly regress a high-resolution disparity map. However, current solutions often suffer from limitations such as the loss of discriminative features caused by downsampling operations that treat all pixels equally, and spatial misalignment resulting from repeated downsampling and upsampling. To overcome these challenges, this paper presents two sampling strategies: the Adaptive Downsampling Module (ADM) and the Disparity Alignment Module (DAM), to prioritize real-time inference while ensuring accuracy. The ADM leverages local features to learn adaptive weights, enabling more effective downsampling while preserving crucial structure information. On the other hand, the DAM employs a learnable interpolation strategy to predict transformation offsets of pixels, thereby mitigating the spatial misalignment issue. Building upon these modules, we introduce ADStereo, a real-time yet accurate network that achieves highly competitive performance on multiple public benchmarks. Specifically, our ADStereo runs over 5x faster than the current state-of-the-art CREStereo (0.054s vs. 0.29s) under the same hardware while achieving comparable accuracy (1.82% vs. 1.69%) on the KITTI stereo 2015 benchmark. The codes are available at: https://***/ cocowy1/ADStereo.

关键词： Depth estimation stereo matching disparity estimation disparity estimation disparity estimation

来源：评论

学校读者我要写书评

暂无评论

Equilibrium image Denoising With Implicit Differentiation

引用

ieee TRANSACTIONS ON image processing 2023年 32卷 1868-1881页

作者： Chen, Qi Wang, Yifei Geng, Zhengyang Wang, Yisen Yang, Jiansheng Lin, Zhouchen Peking Univ Sch Math Sci Beijing 100871 Peoples R China Peking Univ Sch Intelligence Sci & Technol Beijing 100871 Peoples R China Peking Univ Sch Intelligence Sci & Technol Natl Key Lab Gen AI Beijing 100871 Peoples R China Peking Univ Inst Artificial Intelligence Beijing 100871 Peoples R China Peng Cheng Lab Shenzhen 518055 Peoples R China

Recent efforts on learning-based image denoising approaches use unrolled architectures with a fixed number of repeatedly stacked blocks. However, due to difficulties in training networks corresponding to deeper layers, simply stacking blocks may cause performance degradation, and the number of unrolled blocks needs to be manually tuned to find an appropriate value. To circumvent these problems, this paper describes an alternative approach with implicit models. To our best knowledge, our approach is the first attempt to model iterative image denoising through an implicit scheme. The model employs implicit differentiation to calculate gradients in the backward pass, thus avoiding the training difficulties of explicit models and elaborate selection of the iteration number. Our model is parameter-efficient and has only one implicit layer, which is a fixed-point equation that casts the desired noise feature as its solution. By simulating infinite iterations of the model, the final denoising result is given by the equilibrium that is achieved through accelerated black-box solvers. The implicit layer not only captures the non-local self-similarity prior for image denoising, but also facilitates training stability and thereby boosts the denoising performance. Extensive experiments show that our model leads to better performances than state-of-the-art explicit denoisers with enhanced qualitative and quantitative results.

关键词： Iterative methods image denoising computational modeling Training Noise reduction Mathematical models Neural networks deep equilibrium models

来源：评论

学校读者我要写书评

暂无评论

Enhancing Contrast in Circular-view Photoacoustic Computed Tomography Systems

Enhancing Contrast in Circular-view Photoacoustic Computed T...

引用

2024 ieee Ultrasonics, Ferroelectrics, and Frequency Control Joint symposium

作者： Hakakzadeh, Soheil Kavehvash, Zahra Mehrmohammadi, Mohammad Sharif Univ Technol Med Farasout Tajhiz Co Res & Dev Tehran Iran Univ Rochester Med Ctr Dept Imaging Sci Rochester NY 14642 USA

ISBN: (纸本)9798350371918;9798350371901

In this study, we propose a novel method to enhance image quality in sparse circular-view photoacoustic computed tomography (PACT) systems. Traditional PACT systems rely on dense arrays of detectors to achieve high-resolution images, but they suffer from high cost and computational complexity. Our proposed method addresses these challenges by reconstructing multiple low-quality images (LQIs) from under-sampled data and using advanced signal processing techniques, including cross-correlation and the coefficient of variation (CV), to generate a stability ratio (SR) that significantly improves image quality and contrast. We evaluated the proposed method using both artery and Derenzo phantoms, with detector arrays of 1024, 512, and 256 elements. The results demonstrated that our approach effectively suppresses artifacts and enhances image quality even with a reduced number of detectors. Quantitatively, the proposed method showed substantial improvements over conventional techniques, with up to 50% reduction in background noise, a 6 dB increase in signal-to-noise ratio (SNR), and a significant enhancement in structural similarity index (SSIM). These findings suggest that the proposed method enables high-quality PACT imaging with fewer detectors, which is essential to develop a cost-effective and efficient biomedical imaging tool.

关键词： Circular-View PACT Sparse Photoacoustic Imaging image Reconstruction Artifact Suppression

来源：评论

学校读者我要写书评

暂无评论

Insights from Hyperparameter Scaling of Online Speech Separation 14

Insights from Hyperparameter Scaling of Online Speech Separa...

引用

14th International symposium on Chinese Spoken Language processing

作者： Zhou, Xin Zhang, Wangyou Li, Chenda Qian, Yanmin Shanghai Jiao Tong Univ Dept Comp Sci & Engn MoE Key Lab Artificial Intelligence Auditory Cognit & Computat Acoust LabAI Inst Shanghai Peoples R China

ISBN: (纸本)9798331516833;9798331516826

With the rapid development of deep learning, a large number of models with excellent performance for speech separation tasks have emerged in the literature. Despite their impressive performance, these models usually cannot be used for online speech separation tasks due to high computational and memory costs. In recent years, some approaches tailored for online speech separation have been developed to achieve state-of-the-art performance while keeping the model lightweight. However, it is often unclear how these models can scale with their architecture-specific hyperparameters as they are under-explored in the literature. To fill this gap, this paper investigates the effect of hyperparameter scaling on the model performance, number of parameters, and computational cost based on two representative online speech separation models, BSRNN and SkiM. Experimental results on the Libri2Mix dataset reveal the relationship between model performance and composite hyperparameter configurations. We also derive the patterns of the number of parameters and computational costs varying with hyperparameters. Based on the results of the analysis, we give suggestions for scaling the two models and summarize two common trends.

关键词： online speech separation deep neural network model scaling

来源：评论

学校读者我要写书评

暂无评论

GLA-ViT: A Global Contextual Attention Based for Remote Sensing image Scene Classification 22

GLA-ViT: A Global Contextual Attention Based for Remote Sens...

引用

22nd ieee International symposium on Parallel and Distributed processing with Applications, ISPA 2024

作者： Yuan, Hao Liu, Kun Peng, Qinghao Zou, Benli Wang, Can Li, Meishu University of Jinan School of Information Science and Engineering Jinan China Shandong Center of Data and Application China High-Resolution Earth Observation System Jinan China

ISBN: (纸本)9798331509712

Remote Sensing image Scene Classification (RSISC) is a critical area of research in remote sensing that is dedicated to categorizing remote sensing images into specific semantic classes based on their content. Despite the Convolutional Neural Networks' (CNNs) effectiveness in feature extraction and classification, they face challenges in exploring contextual information necessary for a comprehensive understanding of remote sensing scenes. To address this limitation, a novel model based on the Transformer architecture has attracted considerable attention from researchers. The Transformer model has shown promise in extracting contextual information within remote sensing (RS) scenes. However, the original Transformer model's performance in RS scene classification has been subpar due to the diverse content types and scales present in remote sensing scenes. To overcome these challenges, a context-based approach for remote sensing image scene classification called GLA-ViT is proposed. The GLA-ViT model integrates local and global self-attention modules to enhance its ability to capture intricate details and contextual information in remote sensing images, while also maintaining computational efficiency. Additionally, GLA-ViT incorporates an improved MFused-MBConv module to enhance the model's recognition capabilities in complex scenes. Experimental results on well-known datasets such as AID, UCM, and NWPU-RESISC45 demonstrate the superior performance of GLA-ViT in remote sensing image scene classification related tasks. © 2024 ieee.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

MCTE: MARRYING CONVOLUTION AND TRANSFORMER EFFICIENTLY FOR END-TO-END MEDICAL image SEGMENTATION 30

MCTE: MARRYING CONVOLUTION AND TRANSFORMER EFFICIENTLY FOR E...

引用

30th ieee International Conference on image processing (ICIP)

作者： Li, Jiuqiang Southwest Jiaotong Univ Sch Comp & Artificial Intelligence Chengdu Peoples R China

ISBN: (纸本)9781728198354

The purpose of the medical image segmentation task is to delineate different organs or lesion regions in the image, which is an important aid for intelligent clinical medical diagnosis. Recent approaches suffer from the inability to obtain reliable attention, are computationally intensive, and do not exploit the relationships between different samples. We marry convolution and Transformer effectively to establish MCTE for medical image segmentation. The proposed MCTE is an end-to-end network based on U-Net with a parallel learning of three types of attention, namely local attention learning with channel and spatial dimensional convolution, global attention learning with smaller computational effort of swin transformer, and external attention learning with two shared memory storing all medical image information. Extensive experimental results on the ACDC and Synapse dataset, which are widely used for the evaluation of medical image segmentation methods, demonstrate that our proposed method exceeds the compared baseline.

关键词： Medical image Segmentation Transformer Convolution Attention Mechanism

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：