• Journal of Fashion Business
• /
• v.24 no.6
• /
• pp.63-79
• /
• 2020

This study analyze consumers' unconscious visual attention to color and images of internet shopping malls by using eye-tracking method. Twenty-nine participants, including 15 females and 14 males, participated. The average ages of the male and female participants were 27.3 years and 27.7 years, respectively. Ten images of five layouts (multi-composition images, single-model images, gender-composed images, videos, and moving banner images) of internet shopping malls were shown on an eye-tracker computer screen. Quantitative analyses of the eye-tracking responses were conducted. SPSS was used to analyze the descriptive characteristics and to conduct an independent-sample t-test, along with an ANOVA. The data analysis showed that the image area generally had the shortest time to first fixation (TFF), the longest duration of fixation (DOF), the highest number of fixations (NOF), and the highest numbers of revisits(NOR).Notably, visual attention towards female models was high among various images. The results can be used to improve credibility and design online shopping layout with a scientific evidence that helps consumers through their purchase decisions.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.3
• /
• pp.853-873
• /
• 2021

In recent years, deep reinforcement learning (DRL) models are enjoying great interest as their success in a variety of challenging tasks. Deep Q-Network (DQN) is a widely used deep reinforcement learning model, which trains an intelligent agent that executes optimal actions while interacting with an environment. This model is well known for its ability to surpass skilled human players across many Atari 2600 games. Although DQN has achieved excellent performance in practice, there lacks a clear understanding of why the model works. In this paper, we present a visual analytics system for understanding deep Q-network in a non-blind matter. Based on the stored data generated from the training and testing process, four coordinated views are designed to expose the internal execution mechanism of DQN from different perspectives. We report the system performance and demonstrate its effectiveness through two case studies. By using our system, users can learn the relationship between states and Q-values, the function of convolutional layers, the strategies learned by DQN and the rationality of decisions made by the agent.

• International conference on construction engineering and project management
• /
• 2015.10a
• /
• pp.317-320
• /
• 2015

Documenting work-in-progress on construction sites using images captured with smartphones, point-and-shoot cameras, and Unmanned Aerial Vehicles (UAVs) has gained significant popularity among practitioners. The spatial and temporal density of these large-scale site image collections and the availability of 4D Building Information Models (BIM) provide a unique opportunity to develop BIM-driven visual analytics that can quickly and easily detect and visualize construction progress deviations. Building on these emerging sources of information this paper presents a pipeline for model-driven visual analytics of construction progress. It particularly focuses on the following key steps: 1) capturing, transferring, and storing images; 2) BIM-driven analytics to identify performance deviations, and 3) visualizations that enable root-cause assessments on performance deviations. The information requirements, and the challenges and opportunities for improvements in data collection, plan preparations, progress deviation analysis particularly under limited visibility, and transforming identified deviations into performance metrics to enable root-cause assessments are discussed using several real world case studies.

• ETRI Journal
• /
• v.36 no.3
• /
• pp.450-459
• /
• 2014

In this paper, we present a new, easy-to-generate system that is capable of creating virtual 3D tree models and simulating a variety of growth processes of a tree from a single, real tree image. We not only construct various tree models with the same trunk through our proposed digital image matting method and skeleton-based abstraction of branches, but we also animate the visual growth of the constructed 3D tree model through usage of the branch age information combined with a scaling factor. To control the simulation of a tree growth process, we consider tree-growing attributes, such as branching orders, branch width, tree size, and branch self-bending effect, at the same time. Other invisible branches and leaves are automatically attached to the tree by employing parametric branch libraries under the conventional procedural assumption of structure having a local self-similarity. Simulations with a real image confirm that our system makes it possible to achieve realistic tree models and growth processes with ease.

• Journal of information and communication convergence engineering
• /
• v.21 no.4
• /
• pp.346-350
• /
• 2023

Trajectory prediction is an essential element for driving autonomous vehicles, and various trajectory prediction models have emerged with the development of deep learning technology. Convolutional neural network (CNN) is the most commonly used neural network architecture for extracting the features of visual images, and the latest models exhibit high performances. This study was conducted to identify an efficient CNN backbone model among the components of deep learning models for trajectory prediction. We changed the existing CNN backbone network of multiple-trajectory prediction models used as feature extractors to various state-of-the-art CNN models. The experiment was conducted using nuScenes, which is a dataset used for the development of autonomous vehicles. The results of each model were compared using frequently used evaluation metrics for trajectory prediction. Analyzing the impact of the backbone can improve the performance of the trajectory prediction task. Investigating the influence of the backbone on multiple deep learning models can be a future challenge.

• Journal of Broadcast Engineering
• /
• v.25 no.2
• /
• pp.143-150
• /
• 2020

Huge amounts of contents are being uploaded every day on various streaming platforms. Among those videos, game and sports videos account for a great portion. The broadcasting companies sometimes create and provide highlight videos. However, these tasks are time-consuming and costly. In this paper, we propose models that automatically predict highlights in games and sports matches. While most previous approaches use visual information exclusively, our models use both audio and visual information, and present a way to understand short term and long term flows of videos. We also describe models that combine GAN to find better highlight features. The proposed models are evaluated on e-sports and baseball videos.

• Archives of design research
• /
• v.9
• /
• pp.101-114
• /
• 1994

This study deals with the issue of actual HCI design embodiment through the investigation on how to interpret and use visual expression means in HCI design in a semantic way. The purpose of this study is to provide the method by which designers can make the user-centered guidelines in HCI design. As a part of the user\ulcornercentered design approaches, this study is based on the understanding of user group types which are divided by his/her extent of exposure to computer especially and is focussed on applying the level of each group's apprehension of visual expression means to the embodiment of HCI design. Major findings of this study are: First, it proposes the 'Visual Tokens Models' as a basic source for the understanding and the embodiment of visual expression means in HCI design; Second, it has examined the correlations between the characteristics of Visual Tokens and user group types that is, naive users, casual users, and expert users; Third, it proposes guidelines for the user-centered embodiment of HCI design in accordance with the correlations.

• Journal of The Korean Association For Science Education
• /
• v.41 no.1
• /
• pp.19-32
• /
• 2021

This study aims to analyze the visual modeling process of pre-service elementary teachers on the reflection of light. The analytical framework was developed from three aspects; coherence, correspondence, and commensurability of the modeling thinking proposed by Halloun (2004). 20 pre-service elementary teachers participated and were randomly paired to observe the reflection of light. They were asked to construct the visual model individually at first and then collaboratively. Comparing personal and cooperative models, the level of correspondence and commensurability in some groups has increased, despite the lack of special educational treatment. In addition, three main features were found in their reasoning process. First, the level of commensurability to apply the law of reflection continued to change fluidly depending on the circumstances and often the verbal and visual explanations did not match. Second, in the process of visual modeling, correspondence was often given priority over commensurability and coherence. Third, in a situation where correspondence and commensurability are at odds with each other, participants resolved this conflict and developed reasoning through review and revision of the auxiliary hypothesis. Several implications have been discussed for effectively guiding visual modeling activities.

• Journal of the Korean Society for Precision Engineering
• /
• v.22 no.9 s.174
• /
• pp.77-84
• /
• 2005

In most vision applications, we are frequently confronted with determining the position of object continuously. Generally, intertwined processes ire needed for target tracking, composed with tracking and control process. Each of these processes can be studied independently. In case of actual implementation we must consider the interaction between them to achieve robust performance. In this paper, the robust real time visual tracking in complex background is considered. A common approach to increase robustness of a tracking system is to use known geometric models (CAD model etc.) or to attach the marker. In case an object has arbitrary shape or it is difficult to attach the marker to object, we present a method to track the target easily as we set up the color and shape for a part of object previously. Robust detection can be achieved by integrating voting-based visual cues. Kalman filter is used to estimate the motion of moving object in 3D space, and this algorithm is tested in a pan/tilt robot system. Experimental results show that fusion of cues and motion estimation in a tracking system has a robust performance.

• Journal of Electrical Engineering and Technology
• /
• v.10 no.4
• /
• pp.1877-1885
• /
• 2015

This paper presents a new Nearest-Neighbors based weighted representation for images and weighted K-Nearest-Neighbors (WKNN) classifier to improve the precision of image classification using the Bag of Visual Words (BOVW) based models. Scale-invariant feature transform (SIFT) features are firstly extracted from images. Then, the K-means++ algorithm is adopted in place of the conventional K-means algorithm to generate a more effective visual dictionary. Furthermore, the histogram of visual words becomes more expressive by utilizing the proposed weighted vector quantization (WVQ). Finally, WKNN classifier is applied to enhance the properties of the classification task between images in which similar levels of background noise are present. Average precision and absolute change degree are calculated to assess the classification performance and the stability of K-means++ algorithm, respectively. Experimental results on three diverse datasets: Caltech-101, Caltech-256 and PASCAL VOC 2011 show that the proposed WVQ method and WKNN method further improve the performance of classification.