检索结果-内蒙古大学图书馆

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Jin Han Yuta Asano Boxin Shi Yinqiang Zheng Imari Sato Graduate School of Information Science and Technology The University of Tokyo National Institute of Informatics National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University National Engineering Research Center of Visual Technology School of Computer Science Peking University

High-fidelity radiance recovery plays a crucial role in scene information reconstruction and understanding. Conventional cameras suffer from limited sensitivity in dynamic range, bit depth, and spectral response, etc. In this paper, we propose to use event cameras with bio-inspired silicon sensors, which are sensitive to radiance changes, to recover precise radiance values. We reveal that, under active lighting conditions, the transient frequency of event signals triggering linearly reflects the radiance value. We propose an innovative method to convert the high temporal resolution of event signals into precise radiance values. The precise radiance values yields several capabilities in image analysis. We demonstrate the feasibility of recovering radiance values solely from the transient event frequency (TEF) through multiple experiments.

关键词：

来源：评论

学校读者我要写书评

暂无评论

DiLiGenRT: A Photometric Stereo Dataset with Quantified Roughness and Translucency

DiLiGenRT: A Photometric Stereo Dataset with Quantified Roug...

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Heng Guo Jieji Ren Feishi Wang Boxin Shi Mingjun Ren Yasuyuki Matsushita School of Artificial Intelligence Beijing University of Posts and Telecommunications School of Mechanical Engineering Shanghai Jiao Tong University National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University National Engineering Research Center of Visual Technology School of Computer Science Peking University AI Innovation Center School of Computer Science Peking University Graduate School of Information Science and Technology Osaka University

ISBN: (数字)9798350353006

ISBN: (纸本)9798350353013

Photometric stereo faces challenges from non-Lambertian reflectance in real-world scenarios. Systematically measuring the reliability of photometric stereo methods in handling such complex reflectance necessitates a real-world dataset with quantitatively controlled reflectances. This paper introduces DiLiGenRT, the first real-world dataset for evaluating photometric stereo methods under quantified reflectances by manufacturing 54 hemispheres with varying degrees of two reflectance properties: Roughness and Transluency, Unlike qualitative and semantic labels, such as “diffuse” and “specular,” that have been used in previous datasets, our quantified dataset allows comprehensive and systematic benchmark evaluations. In addition, it facilitates selecting best-fit photometric stereo methods based on the quantitative reflectance properties. Our dataset and benchmark results are available at https://***/***.

关键词： Reflectivity computer vision Systematics Semantics Benchmark testing Pattern recognition Manufacturing

来源：评论

学校读者我要写书评

暂无评论

An Efficient Statistical Clock Skew Analysis Method for Clock Trees

An Efficient Statistical Clock Skew Analysis Method for Cloc...

引用

Electronics Design Automation (ISEDA), International Symposium of

作者： Ziyin Cui Tao Zhang Yihui Cai Peng Cao Ting-Jung Lin Lei He National ASIC System Engineering Technology Research Center Southeast University Nanjing China Ningbo Institute of Digital Twin Eastern Institute of Technology Ningbo China Electrical and Computer Engineering Dept University of California Los Angeles CA USA

ISBN: (数字)9798350352030

ISBN: (纸本)9798350352047

Increasing process variability poses great challenge in 2D/3D high-performance clock network design. The variation of device, interconnect and TSV lead to the clock skew variation. However, in most prior models, the variation was not considered comprehensively, where the effect of the equivalent capacitance variation of the interconnect/TSV on the delay of the device was ignored, leading to poor accuracy for clock tree delay variation. In this work, an efficient statistical clock skew analysis method for clock trees is proposed, which considers the variations of device, interconnect, and TSV in 2D/3D clock trees together, and considers the cap variation impact on device variation during the bottom up propagation algorithm. The proposed model was validated under TSMC 22nm process by 3D clock trees implemented for artificial H-tree constructed cases and ISCAS'89 benchmarks. Our model demonstrates excellent agreement with golden Monte Carlo simulation results in terms of the standard deviation of maximum clock skew with the average error of 2.47% while achieves 1400 times speed up. Comparing with competitive works, our model achieves 1.9 times accuracy improvement with comparative simulation effort.

关键词： Solid modeling Accuracy Monte Carlo methods Three-dimensional displays Simulation SPICE Delays

来源：评论

学校读者我要写书评

暂无评论

Friend or foe? Using eye-tracking technology to investigate the visual discrimination ability of giant pandas

引用

Current Zoology 2024年第4期70卷 430-439页

作者： Xinrui Huang Guo Li Guiquan Zhang Zixiang Li Lin Zhao Mengdie Zhu Qinghua Xiang Xuefeng Liu Mei Tian Hemin Zhang Christina D.Buesching Dingzhen Liu Ministry of Education Key Laboratory of Biodiversity Science and Ecological EngineeringDepartment of EcologyCollege of Life SciencesBeijing Normal UniversityBeijing 100875China Key Laboratory of State Forestry and Grassland Administration on Conservation Biology of Rare Animals in the Giant Panda National Park China Conservation and Research Center for the Giant PandaSichuan 623004China School of Computer and Information Technology Beijing Jiaotong UniversityBeijing 100044China Beijing Key Laboratoryof CaptiveWildlife Technology Beijing ZooBeijing100044China Department of Biology Irving K.Barber Faculty of ScienceThe University of British ColumbiaOkanagan CampusKelownaBCCanada

The role that visual discriminative ability plays among giant pandas in social communication and individual discrimination has received less attention than olfactory and auditory ***,we used an eye-tracker technology to investigate pupil fixation patterns for 8 captive male giant pandas Ailuropoda *** paired images(N=26)of conspecifics against:1)sympatric predators(gray wolves and tigers),and non-threatening sympatric species(golden pheasant,golden snub-nosed monkey,takin,and red panda),2)conspecifics with atypical fur colora-tion(albino and brown),and 3)zookeepers/non-zookeepers wearing either work uniform or plain *** each session,we tracked the pan-da's pupil movements and measured pupil first fixation point(FFP),fixation latency,total fixation count(TFC),and duration(TFD)of attention to each ***,pandas exhibited similar attention(FFPs and TFCs)to images of predators and non-threatening sympatric *** of golden pheasant,snub-nosed monkey,and tiger received less attention(TFD)than images of conspecifics,whereas images of takin and red panda received more attention,suggesting a greater alertness to habitat or food competitors than to potential ***'TFCs were greater for images of black-white conspecifics than for albino or brown phenotypes,implying that familiar color elicited more *** reacted differently to images of men versus *** images of women only,pandas gave more attention(TFC)to familiar combinations(uniformed zookeepers and plain-clothed non-zookeepers),consistent with the familiarity *** pandas can use visual perception to discriminate intra-specifically and inter-specifically,including details of human appearance,has applications for panda conservation and captive husbandry.

关键词： eye movement tracker giant panda images total fixation duration total pupil fixation count visual discrimination.

来源：评论

学校读者我要写书评

暂无评论

Quality-Aware End-to-End Audio-Visual Neural Speaker Diarization

arXiv

引用

arXiv 2024年

作者： He, Mao-Kui Du, Jun Niu, Shu-Tong Liu, Qing-Feng Lee, Chin-Hui National Engineering Research Center of Speech and Language Information Processing University of Science and Technology of China Anhui Hefei China IFlytek Hefei Anhui China School of Electrical and Computer Engineering Georgia Institute of Technology AtlantaGA United States

In this paper, we propose a quality-aware end-to-end audio-visual neural speaker diarization framework, which comprises three key techniques. First, our audio-visual model takes both audio and visual features as inputs, utilizing a series of binary classification output layers to simultaneously identify the activities of all speakers. This end-to-end framework is meticulously designed to effectively handle situations of overlapping speech, providing accurate discrimination between speech and non-speech segments through the utilization of multimodal information. Next, we employ a quality-aware audio-visual fusion structure to address signal quality issues for both audio degradations, such as noise, reverberation and other distortions, and video degradations, such as occlusions, off-screen speakers, or unreliable detection. Finally, a cross attention mechanism applied to multi-speaker embedding empowers the network to handle scenarios with varying numbers of speakers. Our experimental results, obtained from various data sets, demonstrate the robustness of our proposed techniques in diverse acoustic environments. Even in scenarios with severely degraded video quality, our system attains performance levels comparable to the best available audio-visual systems. Copyright © 2024, The Authors. All rights reserved.

关键词： Reverberation

来源：评论

学校读者我要写书评

暂无评论

research on Tibetan Part-of-Speech Tagging Based on Transformer 3

Research on Tibetan Part-of-Speech Tagging Based on Transfor...

引用

3rd International Conference on Pattern Recognition and Machine Learning, PRML 2022

作者： Xiangxiu, Cairang Qun, Nuo Renqing, Nuobu Nyima, Trashi Zhao, Qijun School of Information Science and Technology Tibet University China Collaborative Innovation Center for Tibet Informatization by Moe and Tibet Autonomous Region National and Local Joint Engineering Research Center for Tibetan Information Technology Tibet University Lhasa China Tibet University Collaborative Innovation Center for Tibet Informatization by Moe and Tibet Autonomous Region School of Information Science and Technology Lhasa China Tibet University National and Local Joint Engineering Research Center for Tibetan Information Technology Tibet University School of Information Science and Technology Lhasa China Sichuan University School of Information Science and Technology Tibet University Biometrics of Research Laboratory College of Computer Science Chengdu China

ISBN: (数字)9781665499507

ISBN: (纸本)9781665499507

Aiming at the key problems of ambiguity such as polysemy and unregistered words in the Tibetan part-of-speech tagging task, and the currently cited recurrent neural network cannot model long Tibetan texts. This paper proposes Tibetan part-of-speech tagging for a hybrid model of ELMo and Transformer_Encoder. Solved the problem of polysemy and unregistered words in Tibetan that were not solved by static word vectors. ELMo is a bidirectional pre-training model that can extract Tibetan word features in both positive and negative directions. The pre-trained word embeddings are combined with the Transformer's self-attention mechanism to extract the semantic features of Tibetan sentences to predict the part-of-speech tags of each word. This method perfects the Tibetan part-of-speech tagging task and achieves about 97% accuracy on the dataset used in this paper. © 2022 IEEE.

关键词： Recurrent neural networks

来源：评论

学校读者我要写书评

暂无评论

Novel efficient and scalable design of full-adder in atomic silicon dangling bonds (ASDB) technology

引用

Physica Scripta 2023年第12期98卷 125408-125408页

作者： Rasmi, Hadi Mosleh, Mohammad Jafari Navimipour, Nima Kheyrandish, Mohammad Department of Computer Engineering Dezful Branch Islamic Azad University Dezful Iran Department of Computer Engineering Tabriz Branch Islamic Azad University Tabriz Iran Department of Computer Engineering Faculty of Engineering and Natural Sciences Kadir Has University Istanbul Turkey Future Technology Research Center National Yunlin University of Science and Technology Douliou64002 Taiwan

Atomic Silicon Dangling Bonds (ASDB) is an advanced emerging nanotechnology to replace CMOS technology;because it allows the designing of circuits with very high-speed and low-density. However, one of the most critical challenges in implementing circuits in ASDB nanotechnology is output stability and possible defects, such as DB omission, DB misalignment, and DB extra deposition, which can be overcome using a suitable designing pattern. Therefore, developing stable and robust structures is considered as one of essential topics in ASDB. This paper first proposes two novel and stable computing circuits, including a three-input majority voter (MV3) and three-input XOR (XOR3);based on triangular and rhombus patterns, respectively. Then, an efficient ASDB full-adder is designed using the suggested MV3 and XOR3 gates. Finally, two and four-bit ripple carry adders are developed using proposed full-adder. Simulation results indicate that the suggested MV3 and XOR3 are superior to previous designs, by more than 80%, 48%, and 9.5%, averagely;in terms of occupied area, energy, and occurrence, respectively. Moreover, the proposed gates are investigated against possible defects, and the results show high stability. © 2023 IOP Publishing Ltd.

关键词： Nanotechnology

来源：评论

学校读者我要写书评

暂无评论

FMCW Radar-Based Enhancing Moving Target Positioning Accuracy in Spectrogram Mapping

FMCW Radar-Based Enhancing Moving Target Positioning Accurac...

引用

International Conference on Electrical engineering/Electronics, computer, Telecommunications and Information technology, ECTI-CON

作者： Nattakarn Shutimarrungson Seksan Eiadkaew Akkarat Boonpoonga Krit Athikulwongse Suramate Chalermwisutkul Department of Business Computer Mahasarakham Business School Mahasarakham University Maha Sarakham Thailand Department of Electrical and Computer Engineering Faculty of Engineering King Mongkut's University of Technology North Bangkok Bangkok Thailand Department of Electrical and Computer Engineering Faculty of Engineering Research Center of Innovation Digital and Electromagnetic Technology (iDEMT) King Mongkut's University of Technology North Bangkok Bangkok Thailand National Electronics and Computer Technology Center National Science and Technology Development Agency Pathumthani Thailand Communication and Smart System Engineering The Sirindhorn International Thai- German Graduate School of Engineering King Mongkut's University of Technology North Bangkok Bangkok Thailand

ISBN: (数字)9798350381559

ISBN: (纸本)9798350381566

In this paper, we propose a methodology aimed to enhancing the accuracy of moving target positioning using FMCW radar within spectrogram mapping. Utilizing a 10 GHz operating frequency, the FMCW radar effectively detects moving targets. However, the interpretation of intermediate frequency (IF) signals posed challenges due to multiple reflections and resulting ambiguous distances between objects. For solving this problem, we propose two steps of solution: employing an enveloping method and normalization of the IF signal. This process can extract maximum amplitude peaks specific to the target. Experimental validation involved detecting a drone moving away from the radar. Comparative analysis pre and post signal processing in spectrogram terms demonstrate the efficacy of our technique. Furthermore, verification of the detection range of radar was conducted using the GSM-015 GNSS speed meter. Results confirm the accuracy of the positioning capabilities of radar, emphasizing the effectiveness of the proposed approach.

关键词： Visualization Accuracy Refining Radar detection Radar Radar antennas Reflection

来源：评论

学校读者我要写书评

暂无评论

EvPlug: Learn a Plug-and-Play Module for Event and Image Fusion

arXiv

引用

arXiv 2023年

作者： Jiang, Jianping Zhou, Xinyu Duan, Peiqi Shi, Boxin National Key Laboratory for Multimedia Information Processing School of Computer Science China National Engineering Research Center of Visual Technology School of Computer Science China National Key Lab of General AI School of Intelligence Science and Technology Peking University China

Event cameras and RGB cameras exhibit complementary characteristics in imaging: the former possesses high dynamic range (HDR) and high temporal resolution, while the latter provides rich texture and color information. This makes the integration of event cameras into middle- and high-level RGB-based vision tasks highly promising. However, challenges arise in multi-modal fusion, data annotation, and model architecture design. In this paper, we propose EvPlug, which learns a plug-and-play event and image fusion module from the supervision of the existing RGB-based model. The learned fusion module integrates event streams with image features in the form of a plug-in, endowing the RGB-based model to be robust to HDR and fast motion scenes while enabling high temporal resolution inference. Our method only requires unlabeled event-image pairs (no pixel-wise alignment required) and does not alter the structure or weights of the RGB-based model. We demonstrate the superiority of EvPlug in several vision tasks such as object detection, semantic segmentation, and 3D hand pose estimation. Copyright © 2023, The Authors. All rights reserved.

关键词： Image fusion

来源：评论

学校读者我要写书评

暂无评论

Learning Event Guided High Dynamic Range Video Reconstruction

Learning Event Guided High Dynamic Range Video Reconstructio...

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Yixin Yang Jin Han Jinxiu Liang Imari Sato Boxin Shi National Key Laboratory for Multimedia Information Processing School of Computer Science Peking University National Engineering Research Center of Visual Technology School of Computer Science Peking University Graduate School of Information Science and Technology The University of Tokyo National Institute of Informatics

Limited by the trade-off between frame rate and exposure time when capturing moving scenes with conventional cameras, frame based HDR video reconstruction suffers from scene-dependent exposure ratio balancing and ghosting artifacts. Event cameras provide an alternative visual representation with a much higher dynamic range and temporal resolution free from the above issues, which could be an effective guidance for HDR imaging from LDR videos. In this paper, we propose a multimodal learning framework for event guided HDR video reconstruction. In order to better leverage the knowledge of the same scene from the two modalities of visual signals, a multimodal representation alignment strategy to learn a shared latent space and a fusion module tailored to complementing two types of signals for different dynamic ranges in different regions are proposed. Temporal correlations are utilized recurrently to suppress the flickering effects in the reconstructed HDR video. The proposed HDRev-Net demonstrates state-of-the-art performance quantitatively and qualitatively for both synthetic and real-world data.

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：