检索结果-内蒙古大学图书馆

Hybrid descriptor definition for content based image classification using fusion of handcrafted features to convolutional neural network features

学校读者我要写书评

暂无评论

International Journal of Information Technology (Singapore) 2021年第4期13卷 1365-1374页

作者： Das, Rik Kumari, Khusbu De, Sourav Manjhi, P.K. Thepade, Sudeep Department of Post Graduate Program in Information Technology Xavier Institute of Social Service Ranchi Jharkhand India Department of Computer Applications Yogoda Satsanga Mahavidyalaya Ranchi Jharkhand India Department of Computer Science & Engineering Cooch Behar Government Engineering College Cooch Behar India Department of Mathematics Vinoba Bhave University Hazaribag Jharkhand India Department of Computer Engineering Pimpri Chinchwad College of Engineering Pune India

The efficacy of content-based image classification is dependent on the richness of the feature vectors extracted from the image data. Traditional feature extraction techniques highlight single low level image characteristics like shape, size, texture, color etc. for feature generation. The process often fails to extract meaningful descriptors since considering a single image characteristic will ignore other rich properties of image contents. Mass adoption of Convolutional Neural Network (CNN) has significantly improved classification performances for content-based image data. Recent literature has documented high level of precision for image classification by carrying out transfer learning with open source pre trained Convolutional Neural Network (CNN). However, the concept of transfer learning experiences major setback in case of limited training data due to overfitting of training instances during finetuning. This work has identified this challenge and has attempted to capture probability distribution of input images to a pre trained Convolution Neural Network by utilizing it as a fixed weight feature extractor. The proposed approach has addressed the overfitting problem by removing the finetuning step involved in transfer learning. Further investigation on robust descriptor definition is carried out by concatenating the pre-trained CNN features to handcrafted features. The hybrid architecture has encouraging outcomes that have outclassed the classification accuracies of state-of-the-art handcrafted techniques. © 2021, Bharati Vidyapeeth's Institute of computer Applications and Management.

关键词： Augmented feature learning Feature fusion Handcrafted features Image classification Pre trained Convolutional Neural Network (CNN) Representation learning

Do humans and Convolutional Neural Networks attend to similar areas during scene classification: Effects of task and image type

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Müller, Romy Dürschmidt, Marcel Ullrich, Julian Knoll, Carsten Weber, Sascha Seitz, Steffen Faculty of Psychology Chair of Engineering Psychology and Applied Cognitive Research TUD Dresden University of Technology Dresden Germany Faculty of Electrical and Computer Engineering Chair of Fundamentals of Electrical Engineering TUD Dresden University of Technology Dresden Germany Faculty of Mathematics and Natural Sciences Department of Computer Science Machine Learning Group Heinrich-Heine-Universität Düsseldorf Düsseldorf Germany

Deep Learning models like Convolutional Neural Networks (CNN) are powerful image classifiers, but what factors determine whether they attend to similar image areas as humans do? While previous studies have focused on technological factors, little is known about the role of factors that affect human attention. In the present study, we investigated how the tasks used to elicit human attention maps interact with image characteristics in modulating the similarity between humans and CNN. We varied the intentionality of human tasks, ranging from spontaneous gaze during categorization over intentional gaze-pointing up to manual area selection. Moreover, we varied the type of image to be categorized, using either singular, salient objects, indoor scenes consisting of object arrangements, or landscapes without distinct objects defining the category. The human attention maps generated in this way were compared to the CNN attention maps revealed by explainable artificial intelligence (Grad-CAM). The influence of human tasks strongly depended on image type: For objects, human manual selection produced maps that were most similar to CNN, while the specific eye movement task has little impact. For indoor scenes, spontaneous gaze produced the least similarity, while for landscapes, similarity was equally low across all human tasks. To better understand these results, we also compared the different human attention maps to each other. Our results highlight the importance of taking human factors into account when comparing the attention of humans and CNN. © 2023, CC BY.

关键词： Eye movements

Multi-granularity textual adversarial attack with behavior cloning

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Chen, Yangyi Su, Jin Wei, Wei Cognitive Computing and Intelligent Information Processing Laboratory School of Computer Science and Technology Huazhong University of Science Tehchnology School of Software Engineering Huazhong University of Science Tehchnology

Recently, the textual adversarial attack models become increasingly popular due to their successful in estimating the robustness of NLP models. However, existing works have obvious deficiencies. (1) They usually consider only a single granularity of modification strategies (e.g. word-level or sentence-level), which is insufficient to explore the holistic textual space for generation;(2) They need to query victim models hundreds of times to make a successful attack, which is highly inefficient in practice. To address such problems, in this paper we propose MAYA, a MultigrAnularitY Attack model to effectively generate high-quality adversarial samples with fewer queries to victim models. Furthermore, we propose a reinforcement-learning based method to train a multi-granularity attack agent through behavior cloning with the expert knowledge from our MAYA algorithm to further reduce the query times. Additionally, we also adapt the agent to attack black-box models that only output labels without confidence scores. We conduct comprehensive experiments to evaluate our attack models by attacking BiLSTM, BERT and RoBERTa in two different black-box attack settings and three benchmark datasets. Experimental results show that our models achieve overall better attacking performance and produce more fluent and grammatical adversarial samples compared to baseline models. Besides, our adversarial attack agent significantly reduces the query times in both attack settings. Our codes are released at https://***/Yangyi-Chen/MAYA. © 2021, CC BY-NC-SA.

关键词： Cloning

Deep peak property learning for efficient chiral molecules ECD spectra prediction

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Li, Hao Long, Da Yuan, Li Tian, Yonghong Wang, Xinchang Mo, Fanyang School of Electronic and Computer Engineering Peking University Shenzhen China -Preferred Program Peking University Shenzhen Graduate School Shenzhen China College of Chemistry and Chemical Engineering Xiamen University Xiamen361005 China Peng Cheng Laboratory Shenzhen China School of Materials Science and Engineering Peking University Beijing China

Chiral molecule assignation is crucial for asymmetric catalysis, functional materials, and the drug industry. The conventional approach requires theoretical calculations of electronic circular dichroism (ECD) spectra, which is time-consuming and costly. To speed up this process, we have incorporated deep learning techniques for the ECD prediction. We first set up a large-scale dataset of Chiral Molecular ECD spectra (CMCDS) with calculated ECD spectra. We further develop the ECDFormer model, a Transformer-based model to learn the chiral molecular representations and predict corresponding ECD spectra with improved efficiency and accuracy. Unlike other models for spectrum prediction, our ECDFormer creatively focused on peak properties rather than the whole spectrum sequence for prediction, inspired by the scenario of chiral molecule assignation. Specifically, ECDFormer predicts the peak properties, including number, position, and symbol, then renders the ECD spectra from these peak properties, which significantly outperforms other models in ECD prediction, Our ECDFormer reduces the time of acquiring ECD spectra from 1-100 hours per molecule to 1.5s. Copyright © 2024, The Authors. All rights reserved.

关键词： Forecasting

Multi-tenant Cloud SaaS Application for a meeting to task transition via deep learning models

学校读者我要写书评

暂无评论

Multi-tenant Cloud SaaS Application for a meeting to task tr...

Artificial Intelligence and Internet of Things (GCAIoT), IEEE Global Conference

作者： Viktor Walter-Tscharf Department of Information Engineering and Computer Science Doctoral Program Industrial Innovation Fondazione Bruno Kessler University of Trento European Institute of Innovation and Technology

ISBN: (纸本)9798350309850

Thought digitalisation and the post pandemic period online meetings increase by 12.9% and number of attendees by 13.5%. According to Zoom, Google Meets, and MS Teams they collectively increased there amount of active users from March 2020 until june 2021 by 253%. Additionally virtual meetings are becoming inefficient, unproductive and unstructured, where results are not documented correctly. The intention is to shift those statistics by revolutionising the way online meeting are held. The paper outlines the technical requirements for a user-friendly web application that allows to reproduce a virtual meeting including agreements, its participants interactions and highlighting relevant sequences, is outlined. The developed AI meeting wizard autonomously documents the discussed content in the form of notes, to-dos tasks and integrates them directly into task management, mail or calendar tools. Reinforcing the focus on the discussion by increased interaction and productivity instead of task management and note taking.

关键词： Technical requirements Productivity Pandemics Software as a service Speech recognition Internet of Things Task analysis

Gate Stack engineering in n-type (−201) $\upbeta-\text{Ga}_{2}\mathrm{O}_{3}$ Transistors

学校读者我要写书评

暂无评论

Gate Stack Engineering in n-type (−201) $\upbeta-\text{Ga}_...

International Conference on Emerging Electronics (ICEE)

作者： Vishal Khandelwal Saravanan Yuvaraja Chuanju Wang Dhanu Chettri Xiaohang Li Advanced Semiconductor Laboratory Electrical and Computer Engineering Program CEMSE division King Abdullah University of Science and Technology (KAUST) Thuwal Saudi Arabia

A systematic study of Al 2 O 3 and SiO 2 / Al 2 O 3 dielectric on p-Ga2O3 is carried out including the effect of the forming gas annealing. Capacitance-voltage (C-V) curve of the Al 2 O 3 /Ga 2 O 3 gate stack shows a large hysteresis and sweep-to-sweep C-V shift compared to SiO 2 /Al 2 O 3 /Ga 2 O 3 case, which suggests SiO2 interlayer reduces the defect density in Al 2 O 3 /Ga 2 O 3 gate stack. In addition, forming gas annealing further suppresses the interface and border traps density in both Al 2 O 3 /Ga 2 O 3 and SiO 2 / Al 2 O 3 /Ga 2 O 3 gate stacks. A thin film transistor (TFT) with annealed SiO 2 / Al 2 O 3 dielectric has an on-off ratio of ~106, a subthreshold swing (SS) of ~ 0.75 V/decade, and a hysteresis width $(\mathrm{V}_{\text{Hy}})$ of~ 0.5 V compared to an on-off ratio of~105, a SS of ~1.2 V/decade, and a VHy of ~2 V in the controlled TFT with unannealed Al 2 O 3 dielectric. The increased on-off ratio by one order, reduced SS by > 400 mV/decade, and $\mathrm{V}_{\text{Hy}}$ by >1.5 V is attributed to decreased interface and border traps in annealed SiO 2 / Al 2 O 3 compared to unannealed Al 2 O 3 /Ga 2 O 3 gate stack. Interface trap density is calculated by $\mathbf{V}_{\text{Hy}}$ of the transistor, which showed a five-time reduction in annealed SiO 2 / Al 2 O 3 /Ga 2 O 3 compared to the unannealed A;2O3/Ga 2 O 3 gate stack. This study suggests that annealed SiO 2 / Al 2 O 3 dielectric stack is promising for $\boldsymbol{\upbeta}-\text{Ga}_2 \mathrm{O}_3$ transistors.

关键词： Silicon compounds Electron traps Annealing Capacitance-voltage characteristics Logic gates Dielectrics Thin film transistors

Optimization of Certainty Factor Algorithm to Overcome Uncertainty in Expert System Identification of Pests and Diseases of Cocoa

学校读者我要写书评

暂无评论

Optimization of Certainty Factor Algorithm to Overcome Uncer...

Electronic and Electrical engineering and Intelligent System (ICE3IS), International Conference on

作者： Rakhmat Dedi Gunawan Imam Ahmad Parjito Harry Anggono Ego Hengki Vernando Firman Jaya Faculty of Engineering and Computer Science Universitas Teknokrat Indonesia Bandarlampung Indonesia Information Technology Education Study Program STKIP PGRI Situbondo Indonesia

ISBN: (纸本)9781665465427

Lampung Province is one of the regions with the highest cocoa production, but many factors can interfere with cocoa production. One of them is the factor of pests and diseases of cocoa that cannot be identified and prevented beforehand. An expert system is a system that adopts expert knowledge or experts in a particular field. However, the diagnosis usually contains uncertainty in the form of answers and statements used. Certainty Factor (CF) algorithm, is one of the algorithms that is able to overcome uncertainty by providing a value for the level of trust from experts and users. However, because the CF value is based on the level of confidence in the symptoms experienced by the user, it creates a problem, namely the consistency of the user's answers. So, it is necessary to make improvements to the CF algorithm, to overcome the inconsistency of user answers through the use of Consistency Ratio (CR). Through the use of CR, inconsistencies are overcome through comparisons between the consistency index and the random index. Based on the results of the accuracy test, the proposed model is able to produce an accuracy rate of 86.67% in the diagnosis of pests and diseases on cocoa plants.

关键词： Electrical engineering Deep learning Uncertainty Production Indexes Expert systems Intelligent systems

The need and status of sea turtle conservation and survey of associated computer vision advances

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Paul, Aditya Jyoti Cognitive Applications Research Lab India Department of Computer Science and Engineering SRM Institute of Science and Technology Kattankulathur Tamil Nadu603203 India

— For over hundreds of millions of years, sea turtles and their ancestors have swum in the vast expanses of the ocean. They have undergone a number of evolutionary changes, leading to speciation and sub-speciation. However, in the past few decades, some of the most notable forces driving the genetic variance and population decline have been global warming and anthropogenic impact ranging from large-scale poaching, collecting turtle eggs for food, besides dumping trash including plastic waste into the ocean. This leads to severe detrimental effects in the sea turtle population, driving them to extinction. This research focusses on the forces causing the decline in sea turtle population, the necessity for the global conservation efforts along with its successes and failures, followed by an in-depth analysis of the modern advances in detection and recognition of sea turtles, involving Machine Learning and computer Vision systems, aiding the conservation efforts. Copyright © 2021, The Authors. All rights reserved.

关键词： Face recognition

A Comprehensive Survey of Foundation Models in Medicine

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Khan, Wasif Leem, Seowung See, Kyle B. Wong, Joshua K. Zhang, Shaoting Fang, Ruogu J. Crayton Pruitt Family Department of Biomedical Engineering Herbert Wertheim College of Engineering University of Florida GainesvilleFL United States Norman Fixel Institute for Neurological Diseases University of Florida GainesvilleFL United States Department of Neurology University of Florida GainesvilleFL United States Department of Computer Science The University of North Carolina at Charlotte CharlotteNC United States Center for Cognitive Aging and Memory McKnight Brain Institute University of Florida GainesvilleFL United States Department of Electrical and Computer Engineering Herbert Wertheim College of Engineering University of Florida GainesvilleFL United States Department of Computer Information Science Engineering Herbert Wertheim College of Engineering University of Florida GainesvilleFL United States

Foundation models (FMs) are large-scale deep learning models trained on massive datasets, often using self-supervised learning techniques. These models serve as a versatile base for a wide range of downstream tasks, including those in medicine and healthcare. FMs have demonstrated remarkable success across multiple healthcare domains. However, existing surveys in this field do not comprehensively cover all areas where FMs have made significant strides. In this survey, we present a comprehensive review of FMs in medicine, focusing on their evolution, learning strategies, flagship models, applications, and associated challenges. We examine how prominent FMs, such as the BERT and GPT families, are transforming various aspects of healthcare, including clinical large language models, medical image analysis, and omics research. Additionally, we provide a detailed taxonomy of FM-enabled healthcare applications, spanning clinical natural language processing, medical computer vision, graph learning, and other biology- and omics- related tasks. Despite the transformative potentials of FMs, they also pose unique challenges. This survey delves into these challenges and highlights open research questions and lessons learned to guide researchers and practitioners. Our goal is to provide valuable insights into the capabilities of FMs in health, facilitating responsible deployment and mitigating associated risks. Copyright © 2024, The Authors. All rights reserved.

关键词： Self-supervised learning