A Multi-Label Classification with an Adversarial-Based Denoising Autoencoder for Medical Image Annotation

2023 | ACM Transactions on Management Information Systems | Citations: 0

Authors: Chai, Yidong; Liu, Hongyan; Xu, Jie; Samtani, Sagar; Jiang, Yuanchun; Liu, Haoxin

Abstract: Medical image annotation aims to automatically describe the content of medical i ... Expand

Abstract: Medical image annotation aims to automatically describe the content of medical images. It helps doctors to understand the content of medical images and make better informed decisions like diagnoses. Existing methods mainly follow the approach for natural images and fail to emphasize the object abnormalities, which is the essence of medical images annotation. In light of this, we propose to transform the medical image annotation to a multi-label classification problem, where object abnormalities are focused directly. However, extant multi-label classification studies rely on arduous feature engineering, or do not solve label correlation issues well in medical images. To solve these problems, we propose a novel deep learning model where a frequent pattern mining component and an adversarial-based denoising autoencoder component are introduced. Extensive experiments are conducted on a real retinal image dataset to evaluate the performance of the proposed model. Results indicate that the proposed model significantly outperforms image captioning baselines and multi-label classification baselines. Collapse

Topics: workstation

Methods: data transformation autoencoder experiment deep learning literature study

Deep Neural Network at Edge: An Exploration of Hyper Parameter Tuning on Plant Seedling Dataset

2023 | Americas Conference on Information Systems | Citations: 0

Authors: Shaheen, Ali; El-Gayar, Omar

Abstract: This study explored different hyperparameters and their outcomes on multiple DN ... Expand

Abstract: This study explored different hyperparameters and their outcomes on multiple DNN architectures by incorporating transfer learning technique on the plant seedling dataset. Optimizers can have different strengths and weaknesses, and their performance may depend on the specific characteristics of the model and the dataset being used. We have observed that the choice of the optimizer is just one of many hyperparameters that can affect the performance of deep neural network (DNN) models. Other hyperparameters, such as the droprate, number of epochs, and batch size, can also have a significant impact. MobileNetV2 demonstrated superior performance while maintaining a smaller model size, making it a highly valuable option for edge devices where size is a crucial factor. Collapse

Topics: Google+ information system use data visualization Python

Methods: experiment deep learning optimization model Keras machine learning

Effects of Online Crowds on Self-Disclosure Behaviors in Online Reviews: A Multidimensional Examination

2022 | Journal of Management Information Systems | Citations: 0

Authors: Choi, HanByeol Stella; Oh, Wonseok; Kwak, Chanhee; Lee, Junyeong; Lee, Heeseok

Abstract: In online environments (i.e., product review sites in our case), consumers are i ... Expand

Abstract: In online environments (i.e., product review sites in our case), consumers are increasingly interacting and socializing with many "strangers" (i.e., online crowds) as well as sharing personal and product information. Drawing from social norms theory, we examine how the multiple aspects of online crowds affect their self-disclosure behaviors as they provide online reviews and investigate the extent to which prior experience moderates this relationship. Our analysis of data from a leading apparel rental site in the United States uncovers that individuals are inclined to conform to the self-disclosure behaviors of a crowd and divulge more personal information as self-disclosure variance within the group increases. Conversely, individuals are more likely to conceal personal information as a review page becomes crowded. The findings reveal that a reviewer's prior experience of writing a review on the website weakens conformity behavior and reduces the effects of crowdedness. The prior experience also positively interacts with self-disclosure variance in a crowd. Based on these results, we present theoretical implications to literature on social norms and privacy, prior experience, and online reviews. This study also has managerial implications for firms interested in content generation by online reviewers and in review systems where user-generated content is essential. Collapse

Topics: social norm online review data breach website privacy

Methods: computational algorithm descriptive statistic instrumental variables estimation statistical hypothesis test qualitative content analysis

Theories: social influence theory theory of bounded rationality

An exploration of the relation between the visual attributes of thumbnails and the view-through of videos: The case of branded video content

2022 | Decision Support Systems | Citations: 0

Authors: Koh, Byungwan; Cui, Fuquan

Abstract: While browsing through an online video platform, potential viewers decide which ... Expand

Abstract: While browsing through an online video platform, potential viewers decide which videos to click and watch based on the information and impression they obtain from thumbnails. Therefore, a thumbnail needs to be able to tell potential viewers what the video is about (i.e., be informative), and at the same time, a thumbnail needs to grab potential viewers' attention (i.e., be visually appealing). Drawing from the elaboration likelihood model (ELM), we take the visual attributes that are relevant to the informativeness of thumbnails (e.g., element complexity and object complexity) as central route cues, take the visual attributes that are relevant to the aes­ thetics of thumbnails (e.g., celebrity endorsement, colorfulness, brightness, and image quality) as peripheral route cues, and explore the relation between these visual attributes of thumbnails and the view-through of videos using 3,745 marketing videos posted by 38 top brands across 4 industries – the automobile, beverage, restaurant, and toy industries – on YouTube. We construct our variables by extracting features from thumbnails using image mining techniques. Our study contributes to the literature on information systems and marketing by enhancing our understanding of the role of the visual attributes of images in persuading viewers in the online environment. Furthermore, our study provides a possible theoretical basis for studies on information processing and decision support that focus on automatic thumbnail selection. Finally, our study provides useful and practical guidelines for designing templates for optimal thumbnails that grab potential viewers' attention and yield more video views. Collapse

Topics: YouTube marketing management application programming interface Microsoft Azure Python

Methods: image processing computational algorithm OpenCV term frequency–inverse document frequency sentiment analysis

Theories: elaboration likelihood model

Microexpressions in digital humans: perceived affect, sincerity, and trustworthiness

2022 | Electronic Markets | Citations: 2

Authors: Tastemirova, Aliya; Schneider, Johannes; Kruse, Leona Chandra; Heinzle, Simon; Brocke, Jan vom

Abstract: Digital agents with human-like characteristics have become ubiquitous in our soc ... Expand

Abstract: Digital agents with human-like characteristics have become ubiquitous in our society and are increasingly relevant in commercial applications. While some of them closely resemble humans in appearance (e.g., digital humans), they still lack many subtle social cues that are important for interacting with humans. Among them are the so-called microexpressions— facial expressions that are short, subtle, and involuntary. We investigate to what extent microexpressions in digital humans influence people's perceptions and decision-making in order to inform the practices of digital human's design. Our two experiments applied four types of microexpressions based on emotion type (happiness and anger) and intensity (normal and extreme). This paper is among the first to design and evaluate microexpressions with different intensity levels in digital humans. In particular, we leverage the possibilities of digitally (re)designing humans and human perception. These possibilities are feasible only in a digital environment, where it is possible to explore various microexpressions beyond real human beings' physical capabilities. Collapse

Topics: decision making artificial intelligence electronic market online learning system Python

Methods: experiment survey design parametric test ANOVA qualitative interview

Theories: uncanny valley theory

Disentangling the Effects of Instructor Credibility Cues in Bolstering Learners’ Engagement with Health Short Videos

2022 | European Conference On Information Systems | Citations: 0

Authors: Yu, Mengli; Jiang, Na; Lim, Eric; Tan, Chee-Wee

Abstract: With the rapid development of mobile short-video platforms, viewers have greate ... Expand

Abstract: With the rapid development of mobile short-video platforms, viewers have greater access to a diversity of health short videos. Due to relatively homogenized content in these health short videos, instructor credibility is becoming a key determinant of learners' engagement with health short videos. Yet, there is a dearth of research that has sought to elucidate the role of instructor credibility in driving learners' engagement. Building on social presence theory, we classified the source of instructor credibility into four constituent components, namely physical, contextual, psychological, and behavioral features. Additionally, we advance a research model to disentangle the effects of these four instructor credibility cues on learners' engagement. The research model will be validated by employing deep learning algorithms to operationalize our focal variables based on data of health short videos harvested from a popular mobile short-video platform in China. Collapse

Topics: application programming interface social presence data quality Python

Methods: computational algorithm deep learning facial recognition system biometric measurement discrete choice experiment

Theories: social presence theory

Counteracting Dark Web Text-Based CAPTCHA with Generative Adversarial Learning for Proactive Cyber Threat Intelligence

2022 | ACM Transactions on Management Information Systems | Citations: 0

Authors: Zhang, Ning; Ebrahimi, Mohammadreza; Li, Weifeng; Chen, Hsinchun

Abstract: Automated monitoring of dark web (DW) platforms on a large scale is the first st ... Expand

Abstract: Automated monitoring of dark web (DW) platforms on a large scale is the first step toward developing proactive Cyber Threat Intelligence (CTI). While there are efficient methods for collecting data from the surface web, large-scale dark web data collection is often hindered by anti-crawling measures. In particular, text-based CAPTCHA serves as the most prevalent and prohibiting type of these measures in the dark web. Text-based CAPTCHA identifies and blocks automated crawlers by forcing the user to enter a combination of hard-to-recognize alphanumeric characters. In the dark web, CAPTCHA images are meticulously designed with additional background noise and variable character length to prevent automated CAPTCHA breaking. Existing automated CAPTCHA breaking methods have difficulties in overcoming these dark web challenges. As such, solving dark web text-based CAPTCHA has been relying heavily on human involvement, which is labor-intensive and time-consuming. In this study, we propose a novel framework for automated breaking of dark web CAPTCHA to facilitate dark web data collection. This framework encompasses a novel generative method to recognize dark web text-based CAPTCHA with noisy background and variable character length. To eliminate the need for human involvement, the proposed framework utilizes Generative Adversarial Network (GAN) to counteract dark web background noise and leverages an enhanced character segmentation algorithm to handle CAPTCHA images with variable character length. Our proposed framework, DW-GAN, was systematically evaluated on multiple dark web CAPTCHA testbeds. DW-GAN significantly outperformed the state-of-the-art benchmark methods on all datasets, achieving over 94.4% success rate on a carefully collected real-world dark web dataset. We further conducted a case study on an emergent Dark Net Marketplace (DNM) to demonstrate that DW-GAN eliminated human involvement by automatically solving CAPTCHA challenges with no more than three attempts. Our research enables the CTI community to develop advanced, large-scale dark web monitoring. We make DW-GAN code available to the community as an open-source tool in GitHub. Collapse

Topics: darknet open source IT security defense web page IT security

Methods: experiment computational algorithm ablation analysis deep learning case study

Identification of Vessels on Inland Waters Using Low-Quality Video Streams

2021 | HICSS | Citations: 0

Authors: Hyla, Tomasz; Wawrzyniak, Natalia

Abstract: Video surveillance can be used to monitor recreational and commercial vessels on ... Expand

Abstract: Video surveillance can be used to monitor recreational and commercial vessels on inland waters. It is difficult to identify vessels on the basis of registration numbers alone, especially when structure and size are not standardized in a way that allows for easy recognition. In this paper, we propose a novel vessel identification method that can work continuously on streams with a high compression rate or visible compression artifacts. This method can identify all types of moving vessels during daylight. It should work with existing video monitoring systems used to monitor inland waters. When a vessel has hull inscriptions, the method can identify that vessel. If hull inscriptions are not present, the method identifies the direction of movement of the vessel. The tests were performed by real-time video playback of prerecorded samples in the test environment. The method correctly identifies 94% of vessels and the performance indicates that it is suitable for practical application. Collapse

Methods: computational algorithm experiment machine learning OpenCV

A Time-Sensitive IoT Data Analysis Framework

2021 | HICSS | Citations: 0

Authors: Korala, Harindu; Georgakopoulos, Dimitrios; Jayaraman, Prem Prakash; Yavari, Ali

Abstract: This paper proposes a Time-Sensitive IoT Data Analysis (TIDA) framework that mee ... Expand

Abstract: This paper proposes a Time-Sensitive IoT Data Analysis (TIDA) framework that meets the time-bound requirements of time-sensitive IoT applications. The proposed framework includes a novel task sizing and dynamic distribution technique that performs the following: 1) measures the computing and network resources required by the data analysis tasks of a timesensitive IoT application when executed on available IoT devices, edge computers and cloud, and 2) distributes the data analysis tasks in a way that it meets the time-bound requirement of the IoT application. The TIDA framework includes a TIDA platform that implements the above techniques using Microsoft’s Orleans framework. The paper also presents an experimental evaluation that validates the TIDA framework’s ability to meet the time-bound requirements of IoT applications in the smart cities domain. Evaluation results show that TIDA outperforms traditional cloud-based IoT data processing approaches in meeting IoT application time-bounds and reduces the total IoT data analysis execution time by 46.96%. Collapse

Topics: internet of things mobile system cloud computing smart city PostgreSQL

Methods: computational algorithm experiment design artifact optimization model OpenCV

Community Dynamics in Smart City Digital Twins: A Computer Vision-based Approach for Monitoring and Forecasting Collective Urban Hazard Exposure

2021 | HICSS | Citations: 0

Authors: Mavrokapnidis, Dimitri; Mohammadi, Neda; Taylor, John

Abstract: Urbanization and the growth of human population are leading to increased complex ... Expand

Abstract: Urbanization and the growth of human population are leading to increased complexities in the interactions of citizens with public spaces, creating cities that must be more responsive to community dynamics. Despite the critical need for community-based city management, current decision-making approaches are often uninformed by the collective behavior of communities across time and in different locations. Here, we introduce a computer vision-based monitoring and forecasting approach for Smart City Digital Twins (SCDT) that enables integrating collective behavior into the spatiotemporal assessments of exposure to urban heat stress. Our results from a pilot case study in the city of Columbus, GA, demonstrate the significance of integrating community dynamics into monitoring and forecasting urban hazard exposure. This ongoing study highlights how SCDT can make community dynamics more accessible to and responsive by city managers. Collapse

Topics: digital twin smart city production management Facebook missing data

Methods: computational algorithm image processing pilot case study TensorFlow wearables based research