检索结果-内蒙古大学图书馆

LaPE: Layer-adaptive Position Embedding for Vision Transformers with Independent Layer Normalization

学校读者我要写书评

暂无评论

LaPE: Layer-adaptive Position Embedding for Vision Transform...

International Conference on computer Vision (ICCV)

作者： Runyi Yu Zhennan Wang Yinhuai Wang Kehan Li Chang Liu Haoyi Duan Xiangyang Ji Jie Chen School of Electronic and Computer Engineering Peking University Shenzhen China AI for Science (AI4S)-Preferred Program Peking University Shenzhen Graduate School China Peng Cheng Laboratory Shenzhen China Department of Automation and BNRist Tsinghua University Beijing China School of Computer Science and Technology Zhejiang University Zhejiang China

Position information is critical for Vision Transformers (VTs) due to the permutation-invariance of self-attention operations. A typical way to introduce position information is adding the absolute Position Embedding (PE) to patch embedding before entering VTs. However, this approach operates the same Layer Normalization (LN) to token embedding and PE, and delivers the same PE to each layer. This results in restricted and monotonic PE across layers, as the shared LN affine parameters are not dedicated to PE, and the PE cannot be adjusted on a per-layer basis. To overcome these limitations, we propose using two independent LNs for token embeddings and PE in each layer, and progressively delivering PE across layers. By implementing this approach, VTs will receive layer-adaptive and hierarchical PE. We name our method as Layer-adaptive Position Embedding, abbreviated as LaPE, which is simple, effective, and robust. Extensive experiments on image classification, object detection, and semantic segmentation demonstrate that LaPE significantly outperforms the default PE method. For example, LaPE improves +1.06% for CCT on CIFAR100, +1.57% for DeiT-Ti on ImageNet-1K, +0.7 box AP and +0.5 mask AP for ViT-Adapter-Ti on COCO, and +1.37 mIoU for tiny Segmenter on ADE20K. This is remarkable considering LaPE only increases negligible parameters, memory, and computational cost.

关键词：

CLINICAL DETERIORATION PREDICTION IN BRAZILIAN HOSPITALS BASED ON ARTIFICIAL NEURAL NETWORKS AND TREE DECISION MODELS

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Yazdanpanah, Hamed Silva, Augusto C.M. Guedes, Murilo Morales, Hugo M.P. Coelho, Leandro dos S. Moro, Fernando G. Department of Research Robô Laura PR Curitiba Brazil Department of Computer Science University of São Paulo SP São Paulo Brazil Industrial and Systems Engineering Graduate Program Pontifical Catholic University of Parana PR Curitiba Brazil Electrical Engineering Graduate Program Federal University of Parana PR Curitiba Brazil

Early recognition of clinical deterioration (CD) has vital importance in patients' survival from exacerbation or death. Electronic health records (EHRs) data have been widely employed in Early Warning Scores (EWS) to measure CD risk in hospitalized patients. Recently, EHRs data have been utilized in Machine Learning (ML) models to predict mortality and CD. The ML models have shown superior performance in CD prediction compared to EWS. Since EHRs data are structured and tabular, conventional ML models are generally applied to them, and less effort is put into evaluating the artificial neural network's performance on EHRs data. Thus, in this article, an extremely boosted neural network (XBNet) is used to predict CD, and its performance is compared to eXtreme Gradient Boosting (XGBoost) and random forest (RF) models. For this purpose, 103,105 samples from thirteen Brazilian hospitals are used to generate the models. Moreover, the principal component analysis (PCA) is employed to verify whether it can improve the adopted models' performance. The performance of ML models and Modified Early Warning Score (MEWS), an EWS candidate, are evaluated in CD prediction regarding the accuracy, precision, recall, F1-score, and geometric mean (G-mean) metrics in a 10-fold cross-validation approach. According to the experiments, the XGBoost model obtained the best results in predicting CD among Brazilian hospitals' data. © 2022, CC BY-NC-SA.

关键词： Machine learning

Preprocessing Techniques for Enhancing Vision-Based Detection System for Tire Sidewall Extraction

学校读者我要写书评

暂无评论

Preprocessing Techniques for Enhancing Vision-Based Detectio...

International Seminar on Research of Information Technology and Intelligent Systems (ISRITI)

作者： Jia Hao Loo Veronica Lestari Jauw Chin Seong Lim Khairul-Rijal Sarmin Nunung Nurul Qomariyah Ahmad Nurul Fajar Department of Mechanical Materials and Manufacturing Engineering University of Nottingham Malaysia Malaysia Continental Tyre Technology Centre Malaysia Petaling Jaya Malaysia Computer Science Department School of Computing and Creative Arts Bina Nusantara University Jakarta Indonesia Information System Management Department BINUS Graduate Program-Master Information Systems Bina Nusantara University Jakarta Indonesia

ISBN: (数字)9798331519643

ISBN: (纸本)9798331519650

This study aims to develop a system for extracting crucial information from tire sidewalls using Optical Character Recognition (OCR). Initially, images of tire were captured manually by smartphone cameras, including Redmi 9T, iPhone 11, and Galaxy S23 Ultra. The captured images are then transferred to a computer for storage. Subsequently, these images were cropped according to the boundaries identified by Hough Circle Transform (HCT). The cropped images were then further pre-processed. During the pre-processing phase, geometrical transformation and image sharpening techniques are applied to enhance the clarity and readability of the text images. The text is then extracted using Google Vision, with the extracted text categorized by size, DOT, brand and pattern. The results indicated that the effectiveness of image pre-processing was constrained by the accuracy of circle detection, which reached a maximum rate of 87.1%. This causes parts of the text to be cut out inaccurately, leading to a suboptimal extraction accuracy of 55.65%. It is also observed that the Redmi 9T camera produced inconsistent results compared to other devices. Specifically, the iPhone 11 and Samsung Galaxy S23 Ultra demonstrated superior extraction accuracies of 69.71% and 66.37%, respectively, whereas the Redmi 9T achieved a lower extraction accuracy of 37.76%.

关键词： Accuracy Optical character recognition Text categorization Lighting Transforms US department of Transportation Cameras Tires Internet Data mining

Giant Deep-Ultraviolet Circular Dichroism in Synthetic Architectures of Ordered Carbon Nanotubes

学校读者我要写书评

暂无评论

Giant Deep-Ultraviolet Circular Dichroism in Synthetic Archi...

2023 Conference on Lasers and Electro-Optics, CLEO 2023

作者： Doumani, Jacques Lou, Minhan Dewey, Oliver Hong, Nina Fan, Jichao Baydin, Andrey Pasquali, Matteo Yomogida, Yohei Yanagi, Kazuhiro Saito, Riichiro Kono, Junichiro Gao, Weilu Department of Electrical and Computer Engineering Rice University HoustonTX United States Applied Physics Graduate Program Smalley-Curl Institute Rice University HoustonTX United States Department of Electrical and Computer Engineering The University of Utah Salt Lake CityUT United States Carbon Hub Rice University HoustonTX United States Department of Chemical and Biomolecular Engineering Rice University HoustonTX United States J.A. Woollam Co. LincolnNE United States The Smalley-Curl Institute Rice University HoustonTX United States Department of Chemistry Rice University HoustonTX United States Department of Materials Science and NanoEngineering Rice University HoustonTX United States Department of Physics and Astronomy Tokyo Metropolitan University Tokyo Japan Department Physics Tohoku University Sendai Japan Department Physics and Astronomy Rice University HoustonTX United States

We demonstrate tunable, giant, and structure-induced deep-ultraviolet circular dichroism in macroscopically chiral assemblies of racemic carbon nanotubes prepared using two approaches: mechanical-rotation-assisted vac... 详细信息

ISBN: (纸本)9781957171258

关键词： Dichroism

A Multi-phase Deep Learning Approach for Predicting HCC Response to TACE using Complete Computed Tomography

学校读者我要写书评

暂无评论

A Multi-phase Deep Learning Approach for Predicting HCC Resp...

Data science and engineering in Healthcare, Medicine and Biology, IEEE EMBS Special Topic Conference on

作者： Shun Cheng Chang Chien Fu Liu Marta Misztal Yi Hsien Hsieh Che Lin Graduate Institute of Communication Engineering National Taiwan University (NTU) Taiwan School of Electronic Engineering and Computer Science Queen Mary University of London UK Department of Electrical Engineering NTU Taiwan Center for Computational and Systems Biology and Center for Biotechnology NTU Taiwan Smart Medicine and Health Informatics Program NTU Taiwan

In this research study, we compare the predictive performance of two advanced deep learning-based models in order to provide a solution to TACE (Transarterial Chemoembolization) response prediction in HCC (Hepatocellular Carcinoma) patients. Using entire abdominal CT scans enabled a broader perspective available for the model, eliminating the need for segmentation during the preprocessing. Making use of both single-phase and multi-phase CT imaging, we have used DenseNet121 and have obtained an accuracy of 80% for the multi-phase *** Relevance: The ability to predict the effectiveness of TACE treatment prior to its administration makes it possible to provide a better decision-making aid for physicians and patients.

关键词：

Autonomous car using CNN deep learning algorithm 2

学校读者我要写书评

暂无评论

Autonomous car using CNN deep learning algorithm

2nd Annual Conference of science and Technology, ANCOSET 2020

作者： Sonata, I. Heryadi, Y. Lukas, L. Wibowo, A. Computer Science Department BINUS Graduate Program - Doctor of Computer Science Bina Nusantara University Jakarta Jl. Kyai H. Syahdan No.9 Jakarta11480 Indonesia Faculty of Engineering Universitas Katolik Indonesia Atma Jaya Jl Jend. Sudirman No.51 Jakarta12930 Indonesia

Autonomous cars have become an interesting discussion in recent years. Several major automotive industries such as Tesla, GM and Nissan are trying to become pioneers in autonomous car technology. Giant technology companies such as Google Waymo, Baidu and Aptiv also developing autonomous car technology. Several technological approaches are carried out in implementing autonomous cars for recognizing surrounding situations, such as radar, lidar, sonar, GPS, and odometry. An automatic control system is used to control navigation based on the data obtained from these sensors. This paper will discuss the use of CNN deep learning algorithm for recognizing the surrounding environment in creating the automatic navigation required by autonomous cars. The system designed will create and learn the data set that will be taken in advance and the learning outcomes will be implemented in an open simulation system. This simulation shows high accuracy in learning to navigate the autonomous car by observing the surrounding environment. © Published under licence by IOP Publishing Ltd.

关键词： Autonomous vehicles

Inverse designing metamaterials with programmable nonlinear functional responses in graph space

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Maurizi, Marco Xu, Derek Wang, Yu-Tong Yao, Desheng Hahn, David Oudich, Mourad Satpati, Anish Bauchy, Mathieu Wang, Wei Sun, Yizhou Jing, Yun Zheng, Xiaoyu Rayne Department of Materials Science and Engineering University of California Berkeley Berkeley94720 United States Department of Computer Science University of California Los Angeles Los Angeles90095 United States Graduate Program in Acoustics Penn State University PA16802 United States Universite’ de Lorraine CNRS Institut Jean Lamour NancyF-54000 France Department of Civil and Environmental Engineering University of California Los Angeles Los Angeles90095 United States

Material responses to static and dynamic stimuli, represented as nonlinear curves, are design targets for engineering functionalities like structural support, impact protection, and acoustic and photonic bandgaps. Three-dimensional metamaterials offer significant tunability due to their internal structure, yet existing methods struggle to capture their complex behavior-to-structure relationships. We present GraphMetaMat, a graph-based framework capable of designing three-dimensional metamaterials with programmable responses and arbitrary manufacturing constraints. Integrating graph networks, physics biases, reinforcement learning, and tree search, GraphMetaMat can target stress-strain curves spanning four orders of magnitude and complex behaviors, as well as viscoelastic transmission responses with varying attenuation gaps. GraphMetaMat can create cushioning materials for protective equipment and vibration-damping panels for electric vehicles, outperforming commercial materials, and enabling the automatic design of materials with on-demand functionalities. Copyright © 2024, The Authors. All rights reserved.

关键词： Metamaterials

Generative AI for Unmanned Vehicle Swarms: Challenges, Applications and Opportunities

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Liu, Guangyuan Van Huynh, Nguyen Du, Hongyang Hoang, Dinh Thai Niyato, Dusit Zhu, Kun Kang, Jiawen Xiong, Zehui Jamalipour, Abbas In Kim, Dong The School of Computer Science and Engineering The Energy Research Institute @ NTU Interdisciplinary Graduate Program Nanyang Technological University Singapore The Department of Electrical Engineering and Electronics University of Liverpool Liverpool L69 3GJ United Kingdom The School of Electrical and Data Engineering University of Technology Sydney Australia The School of Computer Science and Engineering Nanyang Technological University Singapore College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics Nanjing China The School of Automation Guangdong University of Technology China The Pillar of Information Systems Technology and Design Singapore University of Technology and Design Singapore The School of Electrical and Information Engineering University of Sydney Australia The College of Information and Communication Engineering Sungkyunkwan University Korea Republic of

With recent advances in artificial intelligence (AI) and robotics, unmanned vehicle swarms have received great attention from both academia and industry due to their potential to provide services that are difficult and dangerous to perform by humans. However, learning and coordinating movements and actions for a large number of unmanned vehicles in complex and dynamic environments introduce significant challenges to conventional AI methods. Generative AI (GAI), with its capabilities in complex data feature extraction, transformation, and enhancement, offers great potential in solving these challenges of unmanned vehicle swarms. For that, this paper aims to provide a comprehensive survey on applications, challenges, and opportunities of GAI in unmanned vehicle swarms. Specifically, we first present an overview of unmanned vehicles and unmanned vehicle swarms as well as their use cases and existing issues. Then, an in-depth background of various GAI techniques together with their capabilities in enhancing unmanned vehicle swarms are provided. After that, we present a comprehensive review on the applications and challenges of GAI in unmanned vehicle swarms with various insights and discussions. Finally, we highlight open issues of GAI in unmanned vehicle swarms and discuss potential research directions. © 2024, CC BY.

关键词： Metadata

Transfer Learning for Keypoint Detection in Low-Resolution Thermal TUG Test Images

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Chen, Wei-Lun Hsieh, Chia-Yeh Kao, Yu-Hsiang Liu, Kai-Chun Peng, Sheng-Yu Tsao, Yu Research Center for Information Technology Innovation Academic Sinica Taiwan Graduate Institute of Electrical Engineering National Taiwan University Taiwan Bachelor’s Program in Medical Informatics and Innovative Applications Fu Jen Catholic University Taiwan Department of Electrical Engineering National Taiwan University Taiwan College of Information and Computer Sciences University of Massachusetts AmherstMA01003 United States Department of Electrical Engineering National Taiwan University of Science of Technology Taiwan

This study presents a novel approach to human keypoint detection in low-resolution thermal images using transfer learning techniques. We introduce the first application of the Timed Up and Go (TUG) test in thermal image computer vision, establishing a new paradigm for mobility assessment. Our method leverages a MobileNetV3-Small encoder and a ViTPose decoder, trained using a composite loss function that balances latent representation alignment and heatmap accuracy. The model was evaluated using the Object Keypoint Similarity (OKS) metric from the COCO Keypoint Detection Challenge. The proposed model achieves better performance with AP, AP50, and AP75 scores of 0.861, 0.942, and 0.887 respectively, outperforming traditional supervised learning approaches like Mask R-CNN and ViTPose-Base. Moreover, our model demonstrates superior computational efficiency in terms of parameter count and FLOPS. This research lays a solid foundation for future clinical applications of thermal imaging in mobility assessment and rehabilitation monitoring. Copyright © 2025, The Authors. All rights reserved.

关键词： Thermography (imaging)