检索结果-内蒙古大学图书馆

Deeper Look at Image Salient Object Detection: Bi-Stream Network With a Small Training dataset

ieee TRANSACTIONS ON MULTIMEDIA 2022年 24卷 73-86页

作者： Wu, Zhenyu Li, Shuai Chen, Chenglizhao Hao, Aimin Qin, Hong Beihang Univ State Key Lab Virtual Real Technol & Syst Beijing 100191 Peoples R China Peng Cheng Lab Shenzhen Peoples R China Qingdao Univ Qingdao 266071 Shandong Peoples R China SUNY Stony Brook Stony Brook NY 11794 USA

Compared with the conventional hand-crafted approaches, the deep learning based ISOD (image salient object detection) models have achieved tremendous performance improvements by training exquisitely crafted fancy networks over large-scale training sets. However, do we really need large-scale training set for ISOD? In this article, we provide a deeper insight into the interrelationship between the ISOD performance and the training data. To alleviate the conventional demands for large-scale training data, we provide a feasible way to construct a novel small-scale training set, which only contains 4 K images. To take full advantage of this new set, we propose a novel bi-stream network consisting of two different feature backbones. Benefit from the proposed gate control unit, this bi-stream network is able to achieve complementary fusion status for its subbranches. To our best knowledge, this is the first attempt to use a small-scale training set to compete with other large-scale ones;nevertheless, our method can still achieve the leading SOTA performance on all tested benchmark datasets. Both the code and dataset are publicly available at https://***/wuzhenyubuaa/TSNet.

关键词： Training data Training Feature extraction Logic gates Object detection visualization Task analysis Bi-stream fusion image salient object detection small-scale training set

来源：评论

学校读者我要写书评

暂无评论

LANDSLIDE AREA MAPPING USING SYNTHETIC APERTURE RADAR (SAR) data: THE CASE OF THE 2018 NAGA LANDSLIDE

LANDSLIDE AREA MAPPING USING SYNTHETIC APERTURE RADAR (SAR) ...

引用

ieee International Geoscience and Remote Sensing symposium (IGARSS)

作者： Mayuga, Pauline Joy Tiongson, Samuel Francisco Abao, John Joewen Mendoza, Jean Mari Pernez, Geoffrey Allen Ramirez, Ryan Univ Santo Tomas Dept Civil Engn Manila 1015 Philippines

ISBN: (数字)9781665427920

ISBN: (纸本)9781665427920

The use of satellite-based remote sensing data to investigate ground surface conditions has recently been growing. Specifically, synthetic aperture radar (SAR) data have successfully been utilized to detect ground surface deformations and damaged area mapping. This study uses Sentinel-1 SAR data to map the 2018 Naga landslide area in Cebu Island in the Philippines. The intensity and coherence difference analyses were conducted to detect surface changes resulting from the landslide. The amplitude- and coherence-based results coincide with the actual landslide area and successfully map ground surface mass movement using pre- and post- failure SAR images. This paper presents the unique advantages and potential of Sentinel-1 SAR data for landslide area mapping in large-scale open-pit mining sites for rapid disaster impact assessment.

关键词： SAR amplitude analysis coherence difference analysis landslide area mapping Naga

来源：评论

学校读者我要写书评

暂无评论

AudioInsight: Online Exploration of large Audio datasets for Musical Acoustics

AudioInsight: Online Exploration of Large Audio Datasets for...

引用

Internet of Sounds, International symposium on the

作者： Michael Starakis Chrisoula Alexandraki Department of Music Technology and Acoustics Hellenic Mediterranean University Rethymnon Greece

ISBN: (数字)9798350366525

ISBN: (纸本)9798350366532

New perspectives in correlating physical and perceptual characteristics of musical instruments have emerged through recent advancements in machine learning and high-performance computing. Driving such correlating processes, for instance through training deep learning models, requires the assembly of large audio datasets and the development of tools for Exploratory data analysis (EDA). This article presents AudioInsight, a web application for the audio-visual exploration of sound datasets for musical acoustics. This application is a by-product of our research on drumhead acoustics, which aims at the computational inference of damping strategies resulting in a desired sound texture. Developed in Python and Dash-Plotly, AudioInsight features an interactive and responsive user interface providing tools for dataset exploration, statistical analysis and visualization of parameter distributions and relevant correlations. Moreover, AudioInsight permits generating dataset clusters and exploring sounds by interactively navigating within their 2D or 3D graphic representations. Future enhancements of this application, allowing musicians and researchers to contribute to musical instrument datasets through crowdsourcing, may serve as an example application within the Internet of Sounds (IoS) ecosystem.

关键词： Training Graphics data analysis Three-dimensional displays Statistical analysis Navigation Instruments High performance computing data visualization Acoustics

来源：评论

学校读者我要写书评

暂无评论

Evaluating the Potential and Realized Impact of data Augmentations

Evaluating the Potential and Realized Impact of Data Augment...

引用

2023 ieee symposium Series on Computational Intelligence, SSCI 2023

作者： Heise, David Bear, Helen L. Lincoln University Science Technology & Mathematics Jefferson CityMO United States Ylb Tech Ltd London United Kingdom

ISBN: (纸本)9781665430654

data augmentations have been shown to improve predictive performance of machine learning models in many domains. Augmentations are typically used to improve classification performance, but augmentations can distort the intrinsic properties of the original data, thus reducing the utility of a model for real-world applications. Because augmentations directly affect the training data, and thus also affect the machine learning models trained with said data, intelligent selection of augmentations is as critical as the selection of input features and other options in the machine learning pipeline. Such an approach will enable greater transferability of trained models from the research lab to products and services. This paper presents two metrics to evaluate the potential and realized impact of data augmentations. The first metric, eff-score, assesses the relative efficacy of prospective data augmentations before model training. To observe augmentation effects on the intrinsic properties of the training data, the second metric, nirvana distance, measures the effect of data augmentations beyond overall predictive performance after model training. These metrics are tested with a well known multi-purpose audio data set and augmentations from the domain of environmental sound scene analysis. The relative eff-scores correlate with classification results from predictive models trained on the augmented data sets, and the distance components of the nirvana distance explain results observed but not previously understood from output confusion matrices. These results demonstrate promise for data-driven, efficient selection of data augmentations whilst exposing previously hidden impacts on machine learning models. Furthermore, since eff-score and nirvana distance are domain-independent, these metrics have widespread applicability. © 2023 ieee.

关键词： audio augmentation evaluation data augmentation eff-score nirvana distance visualization

来源：评论

学校读者我要写书评

暂无评论

Leveraging large Language Models for Metagenomic analysis

Leveraging Large Language Models for Metagenomic Analysis

引用

2023 ieee Signal Processing in Medicine and Biology symposium, SPMB 2023

作者： Refahi, M.S. Sokhansanj, B.A. Rosen, G.L. Drexel University Department of Electrical & Computer Engineering PhiladelphiaPA United States

ISBN: (纸本)9798350341256

Analyzing sequencing data from microbiome experiments is challenging, since samples can contain tens of thousands of unique taxa (and their genes) and populations of millions of cells. Reducing the dimensionality of metagenomic data is a crucial step in improving the interpretability of complex genetic information, as metagenomic datasets typically encompass a wide range of genetic diversity and *** this study, we implement RoBERTa, a state-of-the-art large language model, and pre-train it on relatively large genomic datasets to obtain a model that can be used to generate embeddings that can help simplify complex metagenomic data sets. The pre-training process enables RoBERTa to capture the inherent characteristics and patterns present in the genomic sequences. We then evaluate the effectiveness of embeddings generated using the pre-trained RoBERTa model in downstream tasks, with a particular focus on taxonomic classification. To assess whether our method can be generalizable, we conduct extensive downstream analysis on three distinct datasets: 16s rRNA, 28s rRNA, and ITS. By utilizing datasets containing 16S rRNA exclusive to bacteria and eukaryotic mitochondria, as well as datasets containing 28S rRNA and ITS specific to eukaryotes (such as fungi), we were able to assess the performance of RoBERTa embeddings across diverse genomic regions. We tune the RoBERTa model through hyperparameter optimization on each dataset. Our results demonstrate that RoBERTa embeddings exhibit promising results in taxonomic classification compared to conventional methods. © 2023 ieee.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

data-Time Tradeoffs for Optimal k-Thresholding Algorithms in Compressed Sensing

Data-Time Tradeoffs for Optimal k-Thresholding Algorithms in...

引用

ieee International symposium on Information Theory (ISIT)

作者： Xu, Jialiang Zhang, Xu Chinese Acad Sci Acad Math & Syst Sci LSEC Beijing Peoples R China

ISBN: (纸本)9781665421607;9781665421591

Optimal k-thresholding algorithms are a class of k-sparse signal recovery algorithms that overcome the shortcomings of traditional hard thresholding algorithms caused by the oscillation of the residual function. In this paper, a novel convergence analysis for optimal k-thresholding algorithms is established, which reveals the data-time tradeoffs of these algorithms. Both the analysis and numerical results demonstrate that when the number of measurements is small, the algorithms cannot converge;when the number of measurements is suitably large, the number of iterations required for successful recovery has a negative correlation with the number of measurements, and the algorithms can achieve linear convergence. Furthermore, the main theorems indicate that the number of measurements required for successful recovery is of the order of k log(n/k), where n is the dimension of the target signal.

关键词： Compressed sensing sparse optimization data-time tradeoffs optimal k-thresholding

来源：评论

学校读者我要写书评

暂无评论

A Review of Single-Source Deep Unsupervised Visual Domain Adaptation

引用

ieee TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 2022年第2期33卷 473-493页

作者： Zhao, Sicheng Yue, Xiangyu Zhang, Shanghang Li, Bo Zhao, Han Wu, Bichen Krishna, Ravi Gonzalez, Joseph E. Sangiovanni-Vincentelli, Alberto L. Seshia, Sanjit A. Keutzer, Kurt Univ Calif Berkeley Dept Elect Engn & Comp Sci Berkeley CA 94720 USA Carnegie Mellon Univ Sch Comp Sci Pittsburgh PA 15213 USA

large-scale labeled training datasets have enabled deep neural networks to excel across a wide range of benchmark vision tasks. However, in many applications, it is prohibitively expensive and time-consuming to obtain large quantities of labeled data. To cope with limited labeled training data, many have attempted to directly apply models trained on a large-scale labeled source domain to another sparsely labeled or unlabeled target domain. Unfortunately, direct transfer across domains often performs poorly due to the presence of domain shift or dataset bias. Domain adaptation (DA) is a machine learning paradigm that aims to learn a model from a source domain that can perform well on a different (but related) target domain. In this article, we review the latest single-source deep unsupervised DA methods focused on visual tasks and discuss new perspectives for future research. We begin with the definitions of different DA strategies and the descriptions of existing benchmark datasets. We then summarize and compare different categories of single-source unsupervised DA methods, including discrepancy-based methods, adversarial discriminative methods, adversarial generative methods, and self-supervision-based methods. Finally, we discuss future research directions with challenges and possible solutions.

关键词： Task analysis data models Adaptation models visualization Training Loss measurement Learning systems Adversarial learning discrepancy-based methods domain adaptation (DA) self-supervised learning (SSL) transfer learning

来源：评论

学校读者我要写书评

暂无评论

21st International symposium on Web and Wireless Geographical Information Systems, W2GIS 2024

21st International Symposium on Web and Wireless Geographica...

引用

21st International symposium on Web and Wireless Geographical Information Systems, W2GIS 2024

ISBN: (纸本)9783031607950

The proceedings contain 16 papers. The special focus in this conference is on Web and Wireless Geographical Information Systems. The topics include: Exploring Spatio-temporal Dynamics: A Historical analysis of Missing Persons data in Mexico, Revealing Patterns and Trends;Assessing and Managing Soil Quality with Geodata: The IQS Project;Publication of Satellite Earth Observations in the Linked Open data Cloud: Experiment Through the TRACES Project;geospatial Webservices and Reproducibility of Research: Challenges and Needs;TAME II: A Modern Geographic Text Annotation Tool;Towards OGC API - Features Centric GIS Applications Controlled by Object Relational Mapping;smooth Building Footprint Aggregation with Alpha Shapes;in Situ visualization of 6DoF Georeferenced Historical Photographs in Location-Based Augmented Reality;Can large Language Models Automatically Generate GIS Reports?;A Digital Twin Architecture for Intelligent Public Transportation Systems: A FIWARE-Based Solution;an analysis of Container Transportation Multiplex Networks from the Perspective of Shipping Company;a Spatial Interaction Model for the Identification of Urban Functional Regions;Enhancing Efficiency and Privacy of Intelligent Public Transportation Systems Through Federated Learning and EdgeAI;towards a Framework for Personalising Leisure Walking Route Recommendations.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Review of the Machine Learning Based Methods Applicable for analysis of Intracellular Calcium Transients Imaging 21

Review of the Machine Learning Based Methods Applicable for ...

引用

21st ieee International symposium on Intelligent Systems and Informatics, SISY 2023

作者： Korenic, Andrej Lorencin, Ivan Tankovic, Nikola Andjus, Pavle Frank, Domagoj Institute for Physiology and Biochemistry University of Belgrade Faculty of Biology Belgrade Serbia Istrian University of Applied Sciences METRIS Research Centre Pula Croatia Juraj Dobrila University of Pula Faculty of Informatics Pula Croatia Univerisity North Department of Computer Science and Informatics Koprivnica Croatia

ISBN: (纸本)9798350343366

This review explores the application of machine learning (ML) in the classification of astrocytic calcium Ca 2+ signaling, a pivotal aspect of cellular biology. Fluorescence imaging has enabled thorough recording of astrocyte activity, but the resulting large, complex time series data sets defy manual analysis, highlighting the need for automated tools. However, existing tools and algorithms designed for neuronal imaging fail to accommodate astrocytic Ca 2+ time series, underlining the need for dedicated tools. The ML models, with their ability to detect intricate, nonlinear patterns and extract significant features from time series data, are ideally suited for analyzing astrocytic Ca 2+ transients. This review presents a comprehensive overview of methods used in ML for time series analysis, covering both the classification of Ca 2+ signals and broader biomedical signals in the time domain and time-frequency representations. Although ML models offer a promising solution for classifying astrocytic Ca 2+ signals, their optimal performance hinges on accurately labeled, high-quality data. An innovative approach is therefore necessary, one that integrates flexible distance representation, temporal features, and an open-ended classification pipeline, thereby enabling analysis of time-dependent changes in signal patterns without reliance on pre-Assigned labels. This review highlights some of the studies that have successfully employed innovative techniques to deliver noteworthy outcomes in time series analysis and classification. © 2023 ieee.

关键词： Frequency domain analysis

来源：评论

学校读者我要写书评

暂无评论

Synthetic data for AI-Powered Ultrafast Laser Based Micro-structuring Method Description 2nd

Synthetic Data for AI-Powered Ultrafast Laser Based Micro-st...

引用

2nd European symposium on Artificial Intelligence in Manufacturing-ESAIM

作者： Blanco-Filgueira, Beatriz Delgado, Tamara Coto, Andrea Gregores Petit, Celine Bruneel, David Romero, Pablo Muinos-Landin, Santiago AIMEN Technol Ctr Porrino 36410 Pontevedra Spain LASEA Seraing 4102 Belgium

ISBN: (纸本)9783031864919;9783031864896;9783031864889

High-power ultrashort pulse lasers (USPLs) represent a sustainable alternative for industrial surface processing. Laser Surface Texturing (LST) allows modification of surface properties such as mechanical, chemical or optical properties. The adoption of USPLs in industry will enable the surface functionalization of large 3D parts through the LST process. However, a real exploitation of USPLs still faces limitations related to high processing resolution, quality, productivity, and the need for expert knowledge in laser micro-structuring strategies. It is crucial to increase maturity, reliability and throughput of LST on a large scale, as well as to provide the industry with the necessary tools for selecting the suitable laser process for each specific application. Achieving these goals requires significant effort in data collection, which is costly in terms of processing times, results analysis and human resources. Additionally, the availability of software simulators for laser micro-structuring is limited, posing challenges for describing USPLs processes through numerical models and methods. In this context, Artificial Intelligence (AI) models represent a unique tool to enhance process understanding and results analysis, providing rapid prediction and visualization capabilities. However, developing AI-models requires vast amounts of data, which are not easily obtainable experimentally, but simulators can play a vital role in this aspect. We present a combined solution based on machine learning methods fed by synthetic data generated by LS-Plume (R) from LASEA for predicting femtosecond laser single beam-based grooves on stainless steel 316L. The methodology for developing such applications is described, and several regression models are compared. The analysis of metrics and prediction error demonstrates that leveraging the synergy between simulation tools and AI-models can be an effective strategy for training AI-models, thereby avoiding the need for a complete, sys

关键词： laser micro-structuring surface texturing ultrashort pulse laser light-matter interaction synthetic data AI-models

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：