检索结果-南通市图书馆

Egocentric Action Anticipation Based on Unsupervised Gaze Estimation

在线全文

学校读者我要写书评

暂无评论

Wuhan University Journal of Natural Sciences 2021年第3期26卷 207-214页

作者： ZHONG Cengsi fang zhijun GAO Yongbin HUANG Bo School of Electronic and Electrical Engineering Shanghai University of Engineering ScienceShanghai 201620China

Gaze information is important for finding region of interest(ROI)which implies where the next action will *** gaze estimation does not work on EPIC-Kitchens for lack of ground *** this paper,we develop an unsupervised gaze estimation method that helps with egocentric action *** adopt gaze map as a feature representation,and input it into a multiple modality network jointly with red-green-blue(RGB),optical flow and object *** explore the method on EGTEA *** estimated gaze map is further optimized with dilation and Gaussian filter,masked onto the original RGB frame and encoded as the important gaze *** results outperform the strong baseline Rolling-Unrolling LSTMs(RULSTM),with top-5 accuracy achieving 34.31%on the seen test set(S1)and 22.07%on unseen test set(S2).The accuracy is improved by 0.58%and 0.87%,respectively.

关键词： egocentric video action anticipation gaze estimation

A Federated Domain Adaptation Algorithm Based on Knowledge Distillation and Contrastive Learning

在线全文

学校读者我要写书评

暂无评论

Wuhan University Journal of Natural Sciences 2022年第6期27卷 499-507页

作者： HUANG fang fang zhijun SHI Zhicai ZHUANG Lehui LI Xingchen HUANG Bo School of Electrical and Electronic Engineering Shanghai University of Engineering ScienceShanghai 201600China

Smart manufacturing suffers from the heterogeneity of local data distribution across parties,mutual information silos and lack of privacy protection in the process of industry chain *** address these problems,we propose a federated domain adaptation algorithm based on knowledge distillation and contrastive *** distillation is used to extract transferable integration knowledge from the different source domains and the quality of the extracted integration knowledge is used to assign reasonable weights to each source domain.A more rational weighted average aggregation is used in the aggregation phase of the center server to optimize the global model,while the local model of the source domain is trained with the help of contrastive learning to constrain the local model optimum towards the global model optimum,mitigating the inherent heterogeneity between local *** experiments are conducted on the largest domain adaptation dataset,and the results show that compared with other traditional federated domain adaptation algorithms,the algorithm we proposed trains a more accurate model,requires fewer communication rounds,makes more effective use of imbalanced data in the industrial area,and protects data privacy.

关键词： federated learning multi-source domain adaptation knowledge distillation contrastive learning

3D Object Detection Based on Vanishing Point and Prior Orientation

在线全文

学校读者我要写书评

暂无评论

Wuhan University Journal of Natural Sciences 2019年第5期24卷 369-375页

作者： GAO Yongbin ZHAO Huaqing fang zhijun HUANG Bo ZHONG Cengsi Department of Electrical and Electronic Engineering Shanghai University of Engineering Science

3D object detection is one of the most challenging research tasks in computer vision. In order to solve the problem of template information dependency of 3D object proposal in the method of 3D object detection based on 2.5D information, we proposed a 3D object detector based on fusion of vanishing point and prior orientation, which estimates an accurate 3D proposal from 2.5D data, and provides an excellent start point for 3D object classification and localization. The algorithm first calculates three mutually orthogonal vanishing points by the Euler angle principle and projects them into the pixel coordinate system. Then, the top edge of the 2D proposal is sampled by the preset sampling pitch, and the first one vertex is taken. Finally, the remaining seven vertices of the 3D proposal are calculated according to the linear relationship between the three vanishing points and the vertices, and the complete information of the 3D proposal is obtained. The experimental results show that this proposed method improves the Mean Average Precision score by 2.7% based on the Amodal3Det method.

关键词： image analysis 3D object detection prior orientation vanishing point Euler angle

Flavonoid extracts from chrysanthemum with appropriate anthocyanins turn blue when exposed to iron ions

在线全文

学校读者我要写书评

暂无评论

Horticultural Plant Journal 2024年第3期10卷 837-852页

作者： Yanfei Li Jiaying Wang Chenfei Lu Zhongman Wang Chengyan Deng Kang Gao Jingjing Li zhijun fang Hao Liu Yan Hong Silan Dai Beijing Key Laboratory of Ornamental Plants Germplasm Innovation&Molecular Breeding National Engineering Research Center for FloricultureBeijingLaboratory of Urban and Rural Ecological EnvironmentKey Laboratory for Genetics and Breeding of Forest Trees and Ornamental Plants of Ministry of EducationSchool of Landscape ArchitectureBeijing Forestry UniversityBeijing 100083China Beijing Greenland Maintenance and Management Service Center Beijing 102211China Beijing North Asia Landscape Co. Beijing 102211China

Although some species that accumulate only cyanidin(Cy)in nature can produce blue flowers through iron ions,there has been no evidence of blue chrysanthemums being generated in this *** study revealed that flavonoid extracts from the ray florets of the chrysanthemum cultivar‘Wandai Fengguang’turned blue when exposed to Fe^(3+).Samples that could turn blue were labeled as CB(Cy-determined blue flowers),while samples that did not turn blue were labeled as CN(Cy-determined non-blue flowers).After a series of experiments,a stable screening system was established using flavonoid extracts containing NaAc buffer at pH 5.5 and a total anthocyanin concentration(TAC)of 30 μmol·L^(-1),and the addition of Fe^(3+)from 0 to 0.25 μmol·L^(-1)allowed for the selection of five CB samples from 39 chrysanthemum *** five CB samples exhibited flower color phenotypes that belonged to Cluster-I with redness(a*)values ranging from 29.03 to 45.99,yellowness(b*)values from-11.31 to 3.77,and brightness(L*)values from 29.07 to ***,the ratio of TAC to total luteolin concentration(TLC)was found to be a critical factor for distinguishing between CB and CN *** realize the desired blue hue in the flavonoid extracts with the participation of Fe^(3+),a TAC to TLC ratio of 2.25 and above is ***,the protoplasts and ray florets of CB samples that turned blue with the involvement of Fe^(2+)showed great potential for cultivating blue chrysanthemums through ferric-anthocyanin ***,this study reveals that blue flowers can be cultivated through the increase in the iron ion concentration,combined with the accumulation of Cy.

关键词： Chrysanthemum×morifolium Cyanidin Copigment Ferric-anthocyanin chelate Blue flower

Clustering face algorithm of Multi-level blocks using high-level feature

在线全文

学校读者我要写书评

暂无评论

International English Education Research 2017年第5期 59-62页

作者： ZHANG Xiangxiang fang zhijun School of Electronic and Electrical Engineering Shanghai University of Engineering Science

The noise data of large-scale messy face picture sets is difficult to remove, and the classification of face images is usually using clustering, the number of clusters is unknown and the estimation of the number is a complicated problem. In this paper, a new algorithm based on the fusion which is face images with depth feature and multi-level blocks clustering is proposed. We cluster face images which is not directly clustering, but the pictures are divided into multiple levels and using feature fusion of many pictures by selecting proper multiple thresholds. Experimental results demonstrate our cluster face algorithm performs well in MS-Celeb-1 Mtl~, CASlA-WebFace and so on.

关键词：

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

基于特征负反馈卷积的点云分析方法

激光与光电子学进展 2022年第12期59卷 93-105页

作者：邓林涛方志军上海工程技术大学电子电气工程学院上海201620

针对不规则性、无序性和稀疏性给点云分析带来的困难与挑战,提出了融合局部信息提取与全局特征推理的点云分析方法。首先,为了更加有效地进行局部点分组,使用结构感知K近邻(KNN)搜索局部邻域点。其次,基于边卷积改进提出一种特征负反馈... 详细信息

针对不规则性、无序性和稀疏性给点云分析带来的困难与挑战,提出了融合局部信息提取与全局特征推理的点云分析方法。首先,为了更加有效地进行局部点分组,使用结构感知K近邻(KNN)搜索局部邻域点。其次,基于边卷积改进提出一种特征负反馈卷积模块,在映射的高维空间中提取更为准确的局部特征。此外,设计了基于注意力机制的全局语义推理模块,通过强调不同区域的分组点来避免潜在的信息冗余,从而全面地获取点云特征。通过在公开的点云数据集ModelNet40和ShapeNet上进行测试,该方法总体分类精度和总体平均交并比分别达到93.8%和86.4%,定量的评估指标以及定性的可视化实验证明了该方法的准确性和鲁棒性。

关键词：图像处理点云分析特征负反馈卷积注意力机制全局语义推理模块

颅内栓塞术后复发动脉瘤不同干预方式疗效分析

在线全文

学校读者我要写书评

暂无评论

中华解剖与临床杂志 2019年第3期24卷 284-289页

作者：方治军张东北京市大兴区人民医院神经外科 102600 首都医科大学附属北京天坛医院神经外科 100070

目的对比不同的二次干预方式对栓塞术后复发动脉瘤的疗效.方法采用回顾性队列研究方法.纳入2010年1月—2016年12月北京市大兴区人民医院神经外科栓塞术后动脉瘤复发患者共101 例,其中男36例,女65例;首次发病年龄15~69(45 ± 11)岁,复... 详细信息

目的对比不同的二次干预方式对栓塞术后复发动脉瘤的疗效.方法采用回顾性队列研究方法.纳入2010年1月—2016年12月北京市大兴区人民医院神经外科栓塞术后动脉瘤复发患者共101 例,其中男36例,女65例;首次发病年龄15~69(45 ± 11)岁,复发年龄16~76(48 ± 11)岁.根据二次干预方式不同分为手术组(24例)、栓塞组(39例)、栓塞+支架组(30例)、保守组(8例),分析患者治疗前后改良Rankin量表(mRS)评分变化、干预后是否存在并发症、干预后是否再复发,判断不同干预方式的治疗效果.结果 4组复发动脉瘤患者的二次干预时年龄、合并高血压或糖尿病、吸烟、动脉瘤破裂发生率、动脉瘤直径及瘤颈宽度、mRS评分变化及并发症发生率比较差异均无统计学意义(P值均>0.05);各组间动脉瘤部位存在统计学差异(χ2 =37.485, P<0.05);二次干预前后mRS比较各组间差异无统计学意义;二次干预后手术组复发0 例(0% )、栓塞组10 例(25.6% ,10/39)、栓塞+支架组3例(10% ,3/30)、保守组2例复发进展(2/8),4组间动脉瘤再复发率比较差异有统计学意义(χ2 =8.986, P<0.05).结论对于复发动脉瘤,开颅夹闭、介入栓塞及支架辅助等均为有效的治疗方式,其中,开颅夹闭较介入治疗具有更低的再复发率;对于保守治疗的复发动脉瘤患者,需要严密随访观察.

关键词：颅内动脉瘤栓塞,治疗性复发放射学,介入

颅内动脉瘤介入栓塞术后二次复发的影响因素分析

在线全文

学校读者我要写书评

暂无评论

中华解剖与临床杂志 2019年第4期24卷 375-379页

作者：方治军张东北京市大兴区人民医院神经外科 102600 首都医科大学附属北京天坛医院神经外科 100070

目的探讨颅内栓塞术后复发动脉瘤介入治疗后再复发的影响因素。方法采用回顾性队列研究方法。纳入2010年1月—2016年12月北京市大兴区人民医院神经外科二次介入栓塞术治疗的复发动脉瘤患者69例,其中男27例、女42例,年龄25~76(48±10)岁... 详细信息

目的探讨颅内栓塞术后复发动脉瘤介入治疗后再复发的影响因素。方法采用回顾性队列研究方法。纳入2010年1月—2016年12月北京市大兴区人民医院神经外科二次介入栓塞术治疗的复发动脉瘤患者69例,其中男27例、女42例,年龄25~76(48±10)岁;二次介入手术后再复发13例,再复发率为18.8%(13/69)。采用单因素分析及Logistic回归分析方法,分析复发颅内动脉瘤介入治疗后再复发的影响因素。结果单因素分析显示,复发组与非复发组患者性别、年龄、合并高血压、合并糖尿病、吸烟、动脉瘤瘤颈宽度、动脉瘤部位、是否支架辅助比较,差异均无统计学意义(P值均>0.05);而两组患者动脉瘤大小、手术前后改良RANKIN量表(mRS)评分变化情况、术后Raymond分级比较,差异均有统计学意义(Z=-3.396、-5.955、-5.825,P值均<0.05)。Logistic回归分析显示:以RaymondⅠ级为参照,RaymondⅡ级(OR=0.116,95%CI 0.025~0.546,P<0.01)和RaymondⅢ级(OR=0.052,95%CI 0.008~0.350,P<0.01)为二次介入栓塞术后动脉瘤再复发的独立危险因素。结论栓塞不完全(RaymondⅡ级和Ⅲ级)是二次介入栓塞术后动脉瘤再复发的独立危险因素;对于再复发的颅内动脉瘤,二次介入术中选择合适的弹簧圈,尽可能栓塞完全无残留,意味着更低的再复发率。

关键词：颅内动脉瘤复发栓塞,治疗性影响因素分析

在线全文

学校读者我要写书评

暂无评论

全局特征及多层次特征聚合的冠脉分割算法

计算机科学与探索 2021年第5期15卷 958-970页

作者：顾佳方志军田方正上海工程技术大学电子电气工程学院上海201600

冠脉计算机断层扫描血管造影(CTA)图像分割在辅助医生判断血管堵塞、血管疾病诊断等许多实际应用中发挥重要作用。针对CTA图像中存在大量噪声和FCN、U-Net、V-Net等经典深度学习算法分割结果不细腻的问题,提出了全局特征及多层次特征聚... 详细信息

冠脉计算机断层扫描血管造影(CTA)图像分割在辅助医生判断血管堵塞、血管疾病诊断等许多实际应用中发挥重要作用。针对CTA图像中存在大量噪声和FCN、U-Net、V-Net等经典深度学习算法分割结果不细腻的问题,提出了全局特征及多层次特征聚合网络。这种新型的网络由全局特征模块、特征融合与V形细化多层次特征聚合模块以及深度监督三部分组成。全局特征模块综合早期和后期特征信息,在融合丰富的细节和语义信息基础上实现对原始CTA图像过滤操作,生成基础特征。细化V形模块在基础特征的基础上生成不同层次的细化特征图,通过聚合不同层次的细化特征图,得到精准冠脉分割图像。此外,在每一个细化V形模块之后加入深度监督机制来避免梯度消失的问题。对提出的方法进行了定量与定性的分析,结果表明,该方法优于主流基线。消融实验也证明了每个模块的有效性。

关键词：冠脉计算机断层扫描血管造影(CTA)分割全局特征细化V形多层次特征聚合深度监督