摘要:In recent years, with the continuous development of neural networks, human gait recognition technology has gradually matured and been widely used in fields such as identity recognition and health assessment. A multimodal cross gait recognition method combining Transformer is proposed based on the existing gait recognition network GaitPart. It is aiming at the problems of low robustness to appearance changes and poor specificity in feature extraction in existing gait recognition models. On the one hand, it introduces gait contour maps and skeleton maps, designs a multimodal cross fusion module to reduce the impact of appearance changes. On the other hand, it combines Transformer with convolutional neural networks to extract multi-scale temporal features, enrich spatiotemporal features, and improve recognition accuracy. The experimental results show that the improved method achieves accuracy of 98.8%, 95.1%, and 84.2% in three scenarios of normal walking, backpacking, and wearing jackets on the CASIA-B dataset, respectively. At the same time, the results have also been improved in different perspectives on the self built gait dataset. A large number of experiments fully demonstrate the superiority and effectiveness of the improved model.
摘要:This paper addresses the increasingly prominent issues of data compliance in cloud computing environments, taking the perspective of cloud service providers to delve deeply into the rational allocation and effective implementation mechanisms of data compliance responsibilities. It begins by examining and synthesizing domestic and international relevant laws and regulations, thereby clarifying the legal responsibility boundaries for cloud service providers during the processes of data processing, storage, transmission, and protection. Subsequently, the article constructs a comprehensive data compliance responsibility allocation model tailored for cloud services, considering the roles and obligations of cloud service providers, cloud clients, and third-party service providers. Finally, the research proposes a multi-faceted fulfillment mechanism encompassing policy formulation, application of technical means, internal management improvements, and enhanced external collaboration. The aim is to provide theoretical guidance and reference for cloud service providers to establish and improve their data compliance management systems through this study.
关键词:Cloud Service Providers;Data Compliance;Responsibility Allocation;Fulfillment Mechanisms;governance system
摘要:This study aims to improve the accuracy of bone age prediction in children. By introducing lightweight VGG16 and CBAM (Convolutional Block Attention Module) attention mechanisms, we designed a lightweight VGG16 network, initialized the model through transfer learning, and extracted key features using CBAM attention mechanisms. Compared with traditional VGG16, ResNet34 and MobileNetV2, the lightweight VGG16 network showed higher predictive performance in the child bone age prediction dataset, with a MAE (Mean Absolute Error) of 5.01 (month). The experimental results and image comparison verified the superiority of the model, and provided reliable support for improving the prediction of children's bone age.
关键词:lightweight VGG16;Attention Mechanism;Bone age prediction;Transfer learning;Deep learning
摘要:The similarity quantification of the three-dimensional cervical disc model and the prosthetic implant is the basis for solving the problems of three-dimensional morphological retrieval of the cervical endplate, the difficulty of prosthesis implantation, and the lack of post-implantation stability. The similarity calculation of 3D cervical disc models mainly faces the following problems: the relative positions of the two 3D models in Euclidean space are difficult to define, the difference in the shape and size of the cervical intervertebral disc is small, and the geometric features of some surfaces cannot be ignored. In order to solve this problem, this paper proposes a three-dimensional similarity algorithm for cervical intervertebral discs, which adjusts the sagittal plane inclination angle of the three-dimensional surface, and then makes sequence slicing processing on the surface to be compared to extract the truncated curve on the surface. The surface similarity comparison problem in three-dimensional space is transformed into two-dimensional space by the plane truncation method, which effectively reduces the complexity of the problem. Finally, experiments were carried out on multiple groups of three-dimensional models of cervical intervertebral discs to verify the effectiveness of the proposed method.
关键词:CT images of cervical spine;cervical intervertebral discs;similarity;Three-dimensional surfaces;Temporal similarity
摘要:In recent years, the scale and total energy consumption of data centers in China are growing continuously, and the green development of data centers has become the industry consensus. Based on the energy consumption analysis of the work of the data center and the construction of the green data center, this paper discusses the problems in the green development of the data center, and puts forward the solution suggestions.
摘要:With the rapid development of software development technologies, low-code platforms have become a new trend in software development, especially prominent in business process-oriented software development. By analyzing the functional requirements and technical architecture of the hotel room management platform, the Join-Cheer Nvwa low-code platform was chosen for software development. The design and implementation of the hotel room management platform demonstrate that low-code platforms accelerate the development cycle, lower the technical threshold, and enhance efficiency and flexibility. It enables developers to build applications more rapidly and meet the ever-changing business needs.