检索结果-内蒙古大学图书馆

19th International Conference on computer Analysis of Images and Patterns, CAIP 2021

作者： Sayahi, Ikbel Jallouli, Malika Mabrouk, Anouar Ben Amar, Chokri Ben Mahjoub, Mohamed Ali LATIS Laboratory of Advanced Technology and Intelligent Systems Université de Sousse Ecole Nationale d’Ingénieurs de Sousse Sousse4023 Tunisia University of SFax Research Groups on Intelligent Machines Laboratory Private National Engineering School of Monastir Monastir Tunisia Department of Mathematics Higher Institute of Applied Mathematics and Computer Science University of Kairouan Kairouan3100 Tunisia Department of Mathematics Faculty of Science University of Tabuk Tabuk Saudi Arabia College of Computers and Information Technology Taif University Taif Saudi Arabia

ISBN: (纸本)9783030891305

Since the release of the first 3D watermarking algorithm in the last decade, 3D meshes applying such an algorithm have been growing up with a diversity of techniques during the embedding of information into meshes. The main objective is always to secure data shared by remote users. The originality of the present work is issued from the use of spherical harmonics to transform the mesh to the frequency domain to be watermarked, and the entropy concept to estimate the optimal order of the reconstruction. Our algorithm includes three parts. The first is the data pre-processing based on a combination of a signature generated by the well-known Secure Hash Algorithm, and a mesh description, subject to a convolutional error correcting code. The second step is the mesh treatment, consisting in extracting spherical harmonics coefficients to be watermarked. Finally, the third part consists in inserting data into coefficients using LSB and quantification method, and next to reconstruct the watermarked mesh. The experimentation of our approach has shown a high insertion rate, while maintaining the mesh quality. It is also robust against the most popular attacks. Our results showed that the present approach improves the existing works. © 2021, Springer Nature Switzerland AG.

关键词： Entropy

来源：评论

学校读者我要写书评

暂无评论

Scalable marginal likelihood estimation for model selection in deep learning

arXiv

引用

arXiv 2021年

作者： Immer, Alexander Bauer, Matthias Fortuin, Vincent Rätsch, Gunnar Khan, Mohammad Emtiyaz Department of Computer Science ETH Zurich Switzerland Max Planck Institute for Intelligent Systems Germany University of Cambridge United Kingdom RIKEN Center for Advanced Intelligence Project Japan

Marginal-likelihood based model-selection, even though promising, is rarely used in deep learning due to estimation difficulties. Instead, most approaches rely on validation data, which may not be readily available. In this work, we present a scalable marginal-likelihood estimation method to select both hyperparameters and network architectures, based on the training data alone. Some hyperparameters can be estimated online during training, simplifying the procedure. Our marginal-likelihood estimate is based on Laplace’s method and Gauss-Newton approximations to the Hessian, and it outperforms cross-validation and manual-tuning on standard regression and image classification datasets, especially in terms of calibration and out-of-distribution detection. Our work shows that marginal likelihoods can improve generalization and be useful when validation data is unavailable (e.g., in nonstationary settings). Copyright © 2021, The Authors. All rights reserved.

关键词： Network architecture

来源：评论

学校读者我要写书评

暂无评论

Environment Prediction from Sparse Samples for Robotic Information Gathering

Environment Prediction from Sparse Samples for Robotic Infor...

引用

IEEE International Conference on robotics and Automation (ICRA)

作者： Jeffrey A. Caley Geoffrey A. Hollinger Computer Science Department Pacific Lutheran University Tacoma Washington Collaborative Robotics and Intelligent Systems (CoRIS) Institute Oregon State University Corvallis Oregon

ISBN: (数字)9781728173955

ISBN: (纸本)9781728173962

Robots often require a model of their environment to make informed decisions. In unknown environments, the ability to infer the value of a data field from a limited number of samples is essential to many robotics applications. In this work, we propose a neural network architecture to model these spatially correlated data fields based on a limited number of spatially continuous samples. Additionally, we provide a method based on biased loss functions to suggest future areas of exploration to minimize reconstruction error. We run simulated robotic information gathering trials on both the MNIST hand written digits dataset and a Regional Ocean Modeling System (ROMS) ocean dataset for ocean monitoring. Our method outperforms Gaussian process regression in both environments for modeling the data field and action selection.

关键词： Robots Data models Oceans Neural networks Convolution Network architecture Logic gates

来源：评论

学校读者我要写书评

暂无评论

Benchmarking unsupervised object representations for video sequences

引用

The Journal of Machine Learning Research 2021年第1期22卷 8253-8313页

作者： Marissa A. Weis Kashyap Chitta Yash Sharma Wieland Brendel Matthias Bethge Andreas Geiger Alexander S. Ecker Institute of Computer Science University of Göttingen Germany and Institute for Theoretical Physics University of Tübingen Germany Department of Computer Science University of Tübingen Germany Institute for Theoretical Physics University of Tübingen Germany Institute for Theoretical Physics University of Tübingen Germany and Bernstein Center for Computational Neuroscience Tübingen Germany Department of Computer Science University of Tübingen Germany and Max Planck Institute for Intelligent Systems Tübingen Germany Institute of Computer Science University of Göttingen Germany and Campus Institute Data Science Göttingen Germany and Max Planck Institute for Dynamics and Self-Organization Göttingen Germany

Perceiving the world in terms of objects and tracking them through time is a crucial prerequisite for reasoning and scene understanding. Recently, several methods have been proposed for unsupervised learning of object-centric representations. However, since these models were evaluated on different downstream tasks, it remains unclear how they compare in terms of basic perceptual abilities such as detection, figure-ground segmentation and tracking of objects. To close this gap, we design a benchmark with four data sets of varying complexity and seven additional test sets featuring challenging tracking scenarios relevant for natural videos. Using this benchmark, we compare the perceptual abilities of four object-centric approaches: ViMON, a video-extension of MONet, based on recurrent spatial attention, OP3, which exploits clustering via spatial mixture models, as well as TBA and SCALOR, which use explicit factorization via spatial transformers. Our results suggest that the architectures with unconstrained latent representations learn more powerful representations in terms of object detection, segmentation and tracking than the spatial transformer based architectures. We also observe that none of the methods are able to gracefully handle the most challenging tracking scenarios despite their synthetic nature, suggesting that our benchmark may provide fruitful guidance towards learning more robust object-centric video representations.

关键词： unsupervised learning object-centric representations benchmark tracking segmentation

来源：评论

学校读者我要写书评

暂无评论

DISCODVT: Generating long text with discourse-aware discrete variational transformer

arXiv

引用

arXiv 2021年

作者： Ji, Haozhe Huang, Minlie Department of Computer Science and Technology Institute for Artificial Intelligence State Key Lab of Intelligent Technology and Systems Beijing National Research Center Information Science and Technology Tsinghua University Beijing100084 China

Despite the recent advances in applying pretrained language models to generate high-quality texts, generating long passages that maintain long-range coherence is yet challenging for these models. In this paper, we propose DISCODVT, a discourse-aware discrete variational Transformer to tackle the incoherence issue. DISCODVT learns a discrete variable sequence that summarizes the global structure of the text and then applies it to guide the generation process at each decoding step. To further embed discourse-aware information into the discrete latent representations, we introduce an auxiliary objective to model the discourse relations within the text. We conduct extensive experiments on two open story generation datasets and demonstrate that the latent codes learn meaningful correspondence to the discourse structures that guide the model to generate long texts with better long-range coherence. © 2021, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Detecting Worker Attention Lapses in Human-Robot Interaction: An Eye Tracking and Multimodal Sensing Study

arXiv

引用

arXiv 2023年

作者： Dai, Zhuangzhuang Park, Jinha Kaszowska, Aleksandra Li, Chen Dept. of Computer Science Engineering and Physical Science School of Engineering and Applied Science Aston University BirminghamB4 7ET United Kingdom SDU Robotics The Maersk Mc Kinney Moller Institute University of Southern Denmark Campusvej 55 Odense5230 Denmark the Department of Electronic Systems Aalborg University Fredrik Bajers Vej 7 B5-215 Aalborg9220 Denmark Department of Materials and Production Aalborg University Fibigerstræde 16 Aalborg9220 Denmark

The advent of industrial robotics and autonomous systems endow human-robot collaboration in a massive scale. However, current industrial robots are restrained in co-working with human in close proximity due to inability of interpreting human agents’ attention. Human attention study is non-trivial since it involves multiple aspects of the mind: perception, memory, problem solving, and consciousness. Human attention lapses are particularly problematic and potentially catastrophic in industrial workplace, from assembling electronics to operating machines. Attention is indeed complex and cannot be easily measured with single-modality sensors. Eye state, head pose, posture, and manifold environment stimulus could all play a part in attention lapses. To this end, we propose a pipeline to annotate multimodal dataset of human attention tracking, including eye tracking, fixation detection, third-person surveillance camera, and sound. We produce a pilot dataset containing two fully annotated phone assembly sequences in a realistic manufacturing environment. We evaluate existing fatigue and drowsiness prediction methods for attention lapse detection. Experimental results show that human attention lapses in production scenarios are more subtle and imperceptible than well-studied fatigue and drowsiness. © 2023, CC BY.

关键词： Eye tracking

来源：评论

学校读者我要写书评

暂无评论

Optimizing the fluidity in thin-section investment casting for vacuum induction molten AZ91D-1 %wt. CaO

引用

Journal of Magnesium and Alloys 2025年

作者： Carneiro, V.H. Lopes, V. Duarte, I. Puga, H. INESC TEC - Institute for Systems and Computer Engineering Technology and Science R. Dr. Roberto Frias Porto4200-465 Portugal Department of Engineering School of Science and Technology - Universidade de Trás-os-Montes e Alto Douro Quinta de Prados Vila Real5001-801 Portugal INL – International Iberian Nanotechnology Laboratory Avenida Mestre José Veiga Braga4715-330 Portugal CMEMS-UMinho Department of Mechanical Engineering University of Minho Campus de Azurém Guimarães4800-058 Portugal TEMA - Centre for Mechanical Technology and Automation Department of Mechanical Engineering University of Aveiro Campus Universitário de Santiago Aveiro3810-193 Portugal LASI - Intelligent Systems Associate Laboratory Campus Azurém Guimarães4800-058 Portugal

This paper discusses an experimental investigation into the fluidity of AZ91D-1 %wt. CaO magnesium melt via induction for thin-section investment casting. Plaster molds with thin spiral cavities (0.5 to 1.5 mm square sections) were designed and manufactured to assess the impact of casting conditions on filling length, as magnesium alloys cause severe melting and melt-mold exothermic reactions, making investment casting challenging. Combinations of traditional Mg-mold reaction mitigation techniques, such as applying a protective mold coating (Yttria) and vacuum, were examined to determine their role in the filling process. The results suggest that when induction is employed to melt reactive alloys, these methods are not always beneficial, as initially thought. Particularly at higher melt temperatures, the combination of Yttria-coated molds with low-pressure vacuum induction significantly reduce fluidity: vacuum induced melt levitation which promotes oxidation with the residual atmosphere;and Yttria-coating cracking due to thermal stress during the mold fabrication slows filling and promotes significant melt-mold reaction. This study shows that best results to investment cast thin-sections are obtained by avoiding both vacuum and protective coatings, providing a viable route for the precision manufacturing of stent biomedical devices. © 2025 Chongqing University

关键词： Investment casting

来源：评论

学校读者我要写书评

暂无评论

Robust Watermarking Approach for 3D Multiresolution Meshes Based on Multi-wavelet Transform, SHA512 and Turbocodes 19th

Robust Watermarking Approach for 3D Multiresolution Meshes B...

引用

19th International Conference on computer Analysis of Images and Patterns, CAIP 2021

作者： Jallouli, Malika Sayahi, Ikbel Mabrouk, Anouar Ben Mahjoub, Mohamed Ali Amar, Chokri Ben LATIS Laboratory of Advanced Technology and Intelligent Systems Université de Sousse Ecole Nationale d’Ingénieurs de Sousse Sousse4023 Tunisia Research Groups on Intelligent Machines Laboratory Private National Engineering School of Monastir University of Sfax Monastir Tunisia Department of Mathematics Higher Institute of Applied Mathematics and Computer Science University of Kairouan Kairouan3100 Tunisia Department of Mathematics Faculty of Science University of Tabuk Tabuk Saudi Arabia College of Computers and Information Technology Taif University Taif Saudi Arabia

ISBN: (纸本)9783030891305

Since the 3D mesh security 3D mesh security has become intellectual property, 3D watermarking algorithms have continued to appear to secure 3D meshes shared by remote users. The originality of this approach is to insert copyright data in multiresolution domain using multi-wavelet transform which allowed us to significantly increase the size of the watermark. Our algorithm includes then two rounds of insertion where each one applies a different type of wavelet transform. The first one undergoes coefficients generated by Haar wavelet transform. As for the second iteration, it inserts data into coefficients resulting from Schauder wavelet transform. The watermark, which is a combination of the mesh description and source mesh signature (generated using SHA512) are encoded using a Turbo-code before being inserted into wavelet coefficients. Embedding data includes a transformation of these later to spherical coordinate system and a modulation steps. Finally watermarked mesh is reconstructed using inverse wavelet transform. The experimentation of our approach has shown a very high insertion rate due to the use of the hybrid insertion domain, while maintaining the mesh quality. Watermarked mesh and extracted data are obtained in real time. Our approach is also robust against the most popular attacks. Our results showed that the present approach improves the existing works. © 2021, Springer Nature Switzerland AG.

关键词： Wavelet transforms

来源：评论

学校读者我要写书评

暂无评论

Machine learning-based inverse kinematics scalability for prismatic tensegrity structural manipulators

引用

robotics and Autonomous systems 2025年 193卷

作者： Aidar Shakerimov Medetkhan Altymbek Koichi Koganezawa Azamat Yeshmukhametov Department of Robotics Engineering School of Engineering and Digital Sciences Nazarbayev University Astana Kazakhstan Institute of Smart Systems and Artificial Intelligence Nazarbayev University Astana Kazakhstan Computer Science Department SDU University Almaty Kazakhstan

Tensegrity structures are gaining attention due to their distinctive features that stem from wire-driven mechanisms and their highly redundant nature. These features include a lightweight framework, improved resistance to impacts, and ability to carry high payloads. Nonetheless, controlling these structures and understanding their movement remain complex challenges. Our research introduces a pioneering control strategy that utilizes some machine learning algorithms (linear regression, ridge regression, and neural network feedforward) to achieve inverse kinematics for prismatic tensegrity manipulators. This approach has been experimentally validated on two different structures, one with a triangular and the other with a quadrangular configuration, each forming a dual-layer setup. Our experimental results indicate that each of the presented algorithms facilitates the approximate inverse kinematics required for the control of the manipulators with average precision error of 2 cm.

关键词：

来源：评论

学校读者我要写书评

暂无评论

EFaR 2023: Efficient Face Recognition Competition

EFaR 2023: Efficient Face Recognition Competition

引用

IEEE International Joint Conference on Biometrics (IJCB)

作者： Jan Niklas Kolf Fadi Boutros Jurek Elliesen Markus Theuerkauf Naser Damer Mohamad Alansari Oussama Abdul Hay Sara Alansari Sajid Javed Naoufel Werghi Klemen Grm Vitomir Štruc Fernando Alonso-Fernandez Kevin Hernandez Diaz Josef Bigun Anjith George Christophe Ecabert Hatef Otroshi Shahreza Ketan Kotwal Sébastien Marcel Iurii Medvedev Bo Jin Diogo Nunes Ahmad Hassanpour Pankaj Khatiwada Aafan Ahmad Toor Bian Yang Fraunhofer Institute for Computer Graphics Research IGD Germany TU Darmstadt Germany Department of Electrical and Computer Engineering Khalifa University Abu Dhabi United Arab Emirates Laboratory for Machine Intelligence Faculty of Electrical Engineering University of Ljubljana Slovenia Halmstad University Sweden Idiap Research Institute Martigny Switzerland École Polytechnique Fédérale de Lausanne (EPFL) Lausanne Switzerland Université de Lausanne (UNIL) Lausanne Switzerland Institute of Systems and Robotics University of Coimbra Coimbra Portugal Department of Information Security and Communication Technology eHealth and Welfare Security Group Norwegian University of Science and Technology Norway

This paper presents the summary of the Efficient Face Recognition Competition (EFaR) held at the 2023 International Joint Conference on Biometrics (IJCB 2023). The competition received 17 submissions from 6 different teams. To drive further development of efficient face recognition models, the submitted solutions are ranked based on a weighted score of the achieved verification accuracies on a diverse set of benchmarks, as well as the deployability given by the number of floating-point operations and model size. The evaluation of submissions is extended to bias, cross-quality, and large-scale recognition benchmarks. Overall, the paper gives an overview of the achieved performance values of the submitted solutions as well as a diverse set of baselines. The submitted solutions use small, efficient network architectures to reduce the computational cost, some solutions apply model quantization. An outlook on possible techniques that are underrepresented in current solutions is given as well.

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：