检索结果-南通市图书馆

DTCC:Multi-level dilated convolution with transformer for weakly-supervised crowd counting

在线全文

学校读者我要写书评

暂无评论

Computational Visual Media 2023年第4期9卷 859-873页

作者： Zhuangzhuang Miao Yong Zhang Yuan Peng Haocheng Peng baocai yin Beijing Key Laboratory of Multimedia and Intelligent Software Technology Beijing Institute of Artificial IntelligenceFaculty of Information TechnologyBeijing University of TechnologyBeijing 100124China Taiji Computer Corporation Ltd. China

Crowd counting provides an important foundation for public security and urban *** to the existence of small targets and large density variations in crowd images,crowd counting is a challenging *** methods usually apply convolution neural networks(CNNs)to regress a density map,which requires annotations of individual persons and ***-supervised methods can avoid detailed labeling and only require counts as annotations of images,but existing methods fail to achieve satisfactory performance because a global perspective field and multi-level information are usually *** propose a weakly-supervised method,DTCC,which effectively combines multi-level dilated convolution and transformer methods to realize end-to-end crowd *** main components include a recursive swin transformer and a multi-level dilated convolution regression *** recursive swin transformer combines a pyramid visual transformer with a fine-tuned recursive pyramid structure to capture deep multi-level crowd features,including global *** multi-level dilated convolution regression head includes multi-level dilated convolution and a linear regression head for the feature extraction *** module can capture both low-and high-level features simultaneously to enhance the receptive *** addition,two regression head fusion mechanisms realize dynamic and mean fusion *** on four well-known benchmark crowd counting datasets(UCF_CC_50,ShanghaiTech,UCF_QNRF,and JHU-Crowd++)show that DTCC achieves results superior to other weakly-supervised methods and comparable to fully-supervised methods.

关键词： crowd counting transformer dilated convolution global perspective field pyramid

An improved description method of the bumpy texture

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science in China(Series F) 2009年第3期52卷 523-528页

作者： GAO RongHua KONG DeHui yin baocai Beijing Municipal Key Laboratory of Multimedia and Intelligent Software Beijing University of Technology Beijing 100124 China

Bump mapping is a texture-based rendering approach for simulating surface details to make its illumination results have three-dimensional effects. The bumpy properties of an object are determined by height maps. But in the process of generating height maps, a problem arises, i.e. to get a correct value of the pixel height, empirical data should be calculated repeatedly, which proves very complicated, and meanwhile the realistic rendering effect is reduced, because the bumpy property is exaggerated in the height map. Therefore, in this paper, we present a method for describing the details of the bumpy texture, where a new concept ＂bumpy map＂ is introduced to replace the height map. Experimental results demonstrate that the bumpy details produced by the ＂bumpy map＂ are more consistent with the original bumpy texture than by the method of height map.

关键词： bump mapping bumpy texture height map bumpy map

Deep 3D reconstruction:methods,data,and challenges

在线全文

学校读者我要写书评

暂无评论

Frontiers of Information Technology & Electronic Engineering 2021年第5期22卷 652-672页

作者： Caixia LIU Dehui KONG Shaofan WANG Zhiyong WANG Jinghua LI baocai yin Beijing Key Laboratory of Multimedia and Intelligent Software Technology Beijing Institute of Artificial IntelligenceFaculty of Information TechnologyBeijing University of TechnologyBeijing 100124China Multimedia Laboratory School of Computer ScienceUniversity of SydneySydney NSW 2006Australia

Three-dimensional(3D)reconstruction of shapes is an important research topic in the fields of computer vision,computer graphics,pattern recognition,and virtual *** 3D reconstruction methods usually suffer from two bottlenecks:(1)they involve multiple manually designed states which can lead to cumulative errors,but can hardly learn semantic features of 3D shapes automatically;(2)they depend heavily on the content and quality of images,as well as precisely calibrated *** a result,it is difficult to improve the reconstruction accuracy of those methods.3D reconstruction methods based on deep learning overcome both of these bottlenecks by automatically learning semantic features of 3D shapes from low-quality images using deep ***,while these methods have various architectures,in-depth analysis and comparisons of them are unavailable so *** present a comprehensive survey of 3D reconstruction methods based on deep ***,based on different deep learning model architectures,we divide 3D reconstruction methods based on deep learning into four types,recurrent neural network,deep autoencoder,generative adversarial network,and convolutional neural network based methods,and analyze the corresponding methodologies ***,we investigate four representative databases that are commonly used by the above methods in ***,we give a comprehensive comparison of 3D reconstruction methods based on deep learning,which consists of the results of different methods with respect to the same database,the results of each method with respect to different databases,and the robustness of each method with respect to the number of ***,we discuss future development of 3D reconstruction methods based on deep learning.

关键词： Deep learning models Three-dimensional reconstruction Recurrent neural network Deep autoencoder Generative adversarial network Convolutional neural network

在线全文

学校读者我要写书评

暂无评论

基于小波空间的可伸缩视频编码

计算机工程与应用 2005年第36期41卷 19-22页

作者：李锌尹宝才北京工业大学多媒体与智能软件技术北京市重点实验室北京100022

文章提出了一种新的下采样技术提高了可伸缩视频编码的性能。可伸缩性是从空间、时间、SNR(质量)三个方面来提供的,同时还保持了高压缩率。其中基本层视频序列是从上一层经过下采样得到的视频序列。一次下采样后,分辨率降为原视频序列的... 详细信息

文章提出了一种新的下采样技术提高了可伸缩视频编码的性能。可伸缩性是从空间、时间、SNR(质量)三个方面来提供的,同时还保持了高压缩率。其中基本层视频序列是从上一层经过下采样得到的视频序列。一次下采样后,分辨率降为原视频序列的1/2,基本层的运动向量与增强层的运动向量有很大的相关性。目前的做法是使用M PEG滤波器来进行下采样,层间的运动向量的关系减弱了。文章提出使用小波变换对输入的增强层的视频序列进行下采样,基本层就是小波空间的低频部分,这样在基本层的运动估计补偿可以通过相位平移在低频子带中预测,减少运动估计残差,同时,增强层的运动向量可以由基本层小波空间中的运动向量给出估计值。

关键词：可伸缩视频编码运动补偿时域滤波器提升小波

AN IMPROVED 3D FACE SYNTHESIS BASED ON MORPHABLE MODEL

在线全文

学校读者我要写书评

暂无评论

AN IMPROVED 3D FACE SYNTHESIS BASED ON MORPHABLE MODEL

FP-Net:基于任意角度单幅人体图像的正面姿态估计

IFIP TC12 WG12.5-Second IFIP Conference on Artificial Intelligence Applications and Innovations(AIAI2005)

作者： Chengzhang Wang baocai yin Qin Shi Yanfeng Sun Beijing Municipal Multimedia and Intelligent Software Key Lab Beijing University of Technology

A novel model matching method based on improved genetic algorithm is presented in this paper to improve efficiency of matching process for 3D face *** method is independent from initial values and more robust than stochastic gradient descent *** genetic algorithm has strong global searching *** and mutation probability are regulated during optimization process to improve precision and convergence speed of the *** results show our new model matching method has good performance on 3D face synthesis.

关键词： Morphable model 3D face synthesis Genetic Algorithm

来源： cnki会议评论

在线全文

cnki会议

学校读者我要写书评

暂无评论

计算机辅助设计与图形学学报 2022年第10期34卷 1604-1612页

作者：陈路飞张勇唐永正尹宝才北京工业大学信息学部北京人工智能研究院北京100124

准确提取人体正面姿态,能够更好地辅助进行行为识别、图像生成和虚拟试衣工作.然而,侧面图像、背面图像存在人体自遮挡、关键点不可见等问题,使正面姿态提取非常困难.因此,设计并实现了基于任意角度单幅人体图像的正面姿态估计网络FP-Ne... 详细信息

准确提取人体正面姿态,能够更好地辅助进行行为识别、图像生成和虚拟试衣工作.然而,侧面图像、背面图像存在人体自遮挡、关键点不可见等问题,使正面姿态提取非常困难.因此,设计并实现了基于任意角度单幅人体图像的正面姿态估计网络FP-Net(frontal pose network).首先,制作了多角度人体图像数据集,为模型设计提供数据支持;其次,为了提高模型预测结果的准确性,设计了基于Anchor pose的回归模块和基于三维姿态的特征融合模块;最后,通过FP-Net实现了任意角度人体图像的正面姿态提取.在BJUT Taichi和CMU Panoptic数据集上采用PCK评价指标进行消融实验,结果表明所提方法有效地提高了人体正面姿态估计的准确率.

关键词：姿态估计行为分析多角度图像

在线全文

学校读者我要写书评

暂无评论

基于面部动画参数抽取的特征跟踪

计算机工程 2005年第19期31卷 184-185,205页

作者：王珂尹宝才马淑燕郭金铜北京工业大学多媒体与智能软件技术实验室北京100022

提出了一种结合CANDIDE模型和面部动画参数空间参数的方法进行面部特征跟踪和动画参数的提取。采取训练的方法得到面部动画参数和人脸图像之间的关系,直接进行人脸特征的跟踪和面部动画参数的抽取。实验结果表明,该特征跟踪算法有较强... 详细信息

提出了一种结合CANDIDE模型和面部动画参数空间参数的方法进行面部特征跟踪和动画参数的提取。采取训练的方法得到面部动画参数和人脸图像之间的关系,直接进行人脸特征的跟踪和面部动画参数的抽取。实验结果表明,该特征跟踪算法有较强的鲁棒性和准确性。

关键词：面部动面参数空间 CANDIDE 特征跟踪

在线全文

学校读者我要写书评

暂无评论

一种无线信道多描述图像编码算法

计算机工程 2006年第18期32卷 228-230页

作者：李锌尹宝才杨猛苏海斌北京工业大学多媒体与智能软件技术北京市重点实验室北京100022

提出了一种新的图像多描述编码算法,给出了多描述分解的算法和规则,并结合运动补偿时间域滤波(MCTF)进行多描述间的编码,针对无线信道的特点提出了多描述间分层的概念,以适应无线信道传输。实验结果表明,提出的空域分层多描述图像编码(s... 详细信息

提出了一种新的图像多描述编码算法,给出了多描述分解的算法和规则,并结合运动补偿时间域滤波(MCTF)进行多描述间的编码,针对无线信道的特点提出了多描述间分层的概念,以适应无线信道传输。实验结果表明,提出的空域分层多描述图像编码(spatial-layermultipleimagecoding,SL-MDIC)算法有效地利用了MCTF编码的特点,提高了中央信道的编码性能,同时对于收到单个描述或部分描述,解码的图像仍能获得较好的重构图像。

关键词：多描述编码图像处理数据丢失运动补偿时域滤波提升小波

ON EXPLICIT BASIS OF BIVARIATE SPLINE SPACE

在线全文

学校读者我要写书评

暂无评论

Analysis in Theory and Applications 1998年第4期14卷 53-65页

作者： baocai yin Wen Gao, Beijing Polytechnic University, ChinaHarbin Institute of Technology, China Department of Computer Science Beijing Polytechnic University Beijing, 100022 PRC Department of Computer Science Harbin Institute of Technology Harbin, 150001 PRC 1. Department of Computer Science Beijing Polytechnic University 100022 Beijing PRC 2. Department of Computer Science Harbin Institute of Technology 150001 Harbin PRC

For a subdivision △ of a region in d-dimensional Euclidean space. we consider computation of dimension and of basis function in spline space S(△)consisting of all C(?)piecewise Polynomial func- tions over△of degree at most k. A computational scheme is presented for computing the dimension and bases of spline space S(?)(△).This scheme based On the Grobner basis algorithm and the smooth co-factor method for computing multivariate spline. For bivariate splines, explicit basis functions of S(△)age obtained for any integer k and r when△is a cross-cut partition.

关键词： APPI ON EXPLICIT BASIS OF BIVARIATE SPLINE SPACE Math