Vol. 46, No. 1,
Jan. 2019
Digital Library
GPU-Based Real-Time Light Source Estimation for Augmented Reality
Soyoung Park, Sunghun Jo, Sungkil Lee
http://doi.org/10.5626/JOK.2019.46.1.1
To render synthetic objects on static images captured from the real world, estimation of illumination from the images and application for rendering the objects is solicited. In this work, we propose a real-time estimation of light sources, when the 3D-reconstructed geometry of an indoor scene and the corresponding textures are available. To restore the high intensity of the light source, we first convert an LDR image into HDR. The converted image is then used to get 2D positions of the light sources by hierarchical division-based sampling technique. Lastly, 3D positions of the light sources are estimated from texture-geometry mapping. Since the sampling technique stores grid areas generated at each stage in mipmap textures, the division of areas is processed on GPU in a parallel way, which makes it work in real-time. Our approach can be used in rendering where explicit positions of the light sources are asked, such as shadows.
Automatic Generation of HTML Code Based on Web Page Sketch
Bada Kim, Sangmin Park, Taeyeon Won, Junyung Heo
http://doi.org/10.5626/JOK.2019.46.1.9
Various studies have been conducted to automatically encode GUI designs in web application development. In the past study, the focus was on object region detection using computer vision and object detection based on deep-learning. The past reported study had the limitations of incorrect detection or no detection of the object. In the present work, two technologies were applied collectively to reduce the limitations of conventional object detection. The computer vision is used for layout detection, and deep-learning is used for GUI object detection. Based on these technologies, detected layouts and GUI objects were converted into HTML code. Consequently, the accuracy and recall rate of GUI object detection were 91% and 86%, respectively, and it was possible to convert into HTML code.
Automatic Transformation of Korean Fonts using Unbalanced U-net and Generative Adversarial Networks
Pangjia, Seunghyun Ko, Yang Fang, Geun-sik Jo
http://doi.org/10.5626/JOK.2019.46.1.15
In this paper, we study the typography transfer problem: transferring a source font, to an analog font with a specified style. To solve the typography transfer problem, we treat the problem as an image-to-image translation problem, and propose an unbalanced u-net architecture based on Generative Adversarial Network(GAN). Unlike traditional balanced u-net architecture, architecture we proposed consists of two subnets: (1) an unbalanced u-net is responsible for transferring specified fonts style to another, while maintaining semantic and structure information; (2) an adversarial net. Our model uses a compound loss function that includes a L1 loss, a constant loss, and a binary GAN loss to facilitate generating desired target fonts. Experiments demonstrate that our proposed network leads to more stable training loss, with faster convergence speed in cheat loss, and avoids falling into a degradation problem in generating loss than balanced u-net.
Korean Dependency Parsing using the Self-Attention Head Recognition Model
http://doi.org/10.5626/JOK.2019.46.1.22
Dependency parsing is the problem solving of structural ambiguities of natural language in sentences. Recently, various deep learning techniques have been applied and shown high performance. In this paper, we analyzed deep learning based dependency parsing problem in three stages. The first stage was a representation step for a word (eojeol) that is a unit of dependency parsing. The second stage was a context reflecting step that reflected the surrounding word information for each word. The last stage was the head word and dependency label recognition step. In this paper, we propose the max-pooling method that is widely used in the CNN model for a word representation. Moreover, we apply the Minimal-RNN Unit that has less computational complexity than the LSTM and GRU for contextual representation. Finally, we propose a Self-Attention Head Recognition Model that includes the relative distance embedding between each word for the head word recognition, and applies multi-task learning to the dependency label recognition simultaneously. For the evaluation, the SEJONG phrase-structure parsing corpus was transformed according to the TTA Standard Dependency Guideline. The proposed model showed the accuracy of parsing for UAS 93.38% and LAS 90.42%.
A Combined Model of Outline Feature Map and CNN for Detection of People at the Beach
http://doi.org/10.5626/JOK.2019.46.1.31
As water safety accidents occur every year, many intelligent video surveillance systems are being developed to prevent water safety accidents. In this paper, we propose InsightCNN to accurately detect moving objects in complex images, such as beaches, in intelligent video surveillance systems. First, a basic model was constructed using 1x1 Convolution of Fully Convolutional Network and Residual Block of ResNet. We added an outline feature map that shows a key feature of the image, to the initial layer of the basic model. Results of the experiment demonstrate superiority of the idea of InsightCNN.
Saliency-based SVG Image Placeholder Generation
http://doi.org/10.5626/JOK.2019.46.1.39
Image placeholders are small file-size images used instead of original images for fast and efficient loading of large file-size images, or large-scale of images on the web. While various image placeholder generation techniques based on Scalar Vector Graphics (SVG) have been proposed, file size of the image placeholder increases inevitably for detailed description. Our method generates an SVG-based image placeholder by optimal placement of geometric primitives based on visual saliency. This method generates a smaller file-size placeholder while showing similar visual effects. From experimental results, we could observe that the saliency-based method produces high level-of-detail without increasing file size. Saliency-based SVG image placeholder generation will be useful for web-based services.
Korean Movie-review Sentiment Analysis Using Parallel Stacked Bidirectional LSTM Model
Yeongtaek Oh, Mintae Kim, Wooju Kim
http://doi.org/10.5626/JOK.2019.46.1.45
The sentiment analysis is a field of document classification that classifies the sensitivity of text documents. The sentiment analysis methodology that employs the use of deep learning can be divided into a process of tokenizing a document, obtaining a sentence vector through embedding and classifying a vectorized document. We reviewed the methods of various existing studies and found out the appropriate methodology focusing on embedding methods and deep learning models for the Korean documents through comparative experiments. The document pre-processing method compares documents to words, syllables and phonemes. Additionally, a comparative experiment was conducted on the Naver movie review data set nsmc (naver sentiment movie corpus) from the CNN to the LSTM, bi-LSTM, stacked bi-LSTM and the newly proposed Parallel Stacked Bidirectional LSTM model. The results showed that the performance of the proposed model was higher than that of the existing simple deep learning model. Moreover, itachieved the best classification performance of approximately 88.95% through the ensemble among the models learned through other pre-processing.
Semantic Relationship between Safety Analysis Techniques to Support Traceability in Developing multiple CPSs
Seungwoo Nam, Horn Daneth, Jang-Eui Hong
http://doi.org/10.5626/JOK.2019.46.1.50
Many systems based on convergence IT are being developed in the form of CPS(Cyber Physical System), in which sensors and actuators interwork in aviation, medical, and industrial manufacturing. It is necessary to analyze and prepare for possible hazards from interaction and environmental factors before system operation in real environment, because the collaboration and common functions may appear to achieve the given mission. This paper proposes three extensions of existing safety analysis techniques, for multiple CPSs in relation to analysis activities of IEC 61508 standard. We also identify and define semantic relationships between our extended techniques, to provide the traceability of functional safety in CPSs, and show the usefulness of relationships, by applying to AIDS(Automatic Incident Detection System) of autonomous vehicles.
Automatic Teeth Separation through Searching Teeth Separation Lines and Planes based on Intensity Cost Function Optimization in Maxillofacial CBCT Images
Soyoung Lee, Min Jin Lee, Helen Hong
http://doi.org/10.5626/JOK.2019.46.1.65
In this paper, we propose an automatic teeth separation method based on the intensity cost function which finds optimal teeth separation line in 2D panorama image. It also finds the optimal teeth separation plane considering the spatial information of teeth in 3D maxillofacial CBCT images. First, to observe the overall structure of an individual tooth, the 2D panorama image of the teeth is reconstructed by extracting the teeth arch curve of the crown and root region. Second, the optimal teeth separation lines are searched through the intensity-based cost function in the reconstructed 2D panorama image. Third, to improve the accuracy of teeth separation, the optimal separation planes considering the spatial information are searched through the intensity-based cost function in 3D CBCT images. Experimental results show that the proposed method improves the separation accuracy compared to the comparative method. The average intensity value of the proposed method was reduced by 8.61% compared to the comparative method. The processing time of the proposed method was completed within 30 seconds.
Diagnostic and Therapeutic Model for Korean Major Depressive Disorder Using Multi-Modal Data
Yonghwa Choi, Aram Kim, Minji Jeon, Sunkyu Kim, Kyu-Man Han, Eunsoo Won, Byung-Joo Ham, Jaewoo Kang
http://doi.org/10.5626/JOK.2019.46.1.71
Depression is one of the most common mental illnesses in the modern society, and it increases the social burden due to repeated recurrences. However, since there are many pre-disposing factors that cause depression, there is need to develop a machine-learning model that examine these factors effectively. In this paper, we propose a model that can diagnose depression and predict the degree of antidepressant response using four multi modal data including basic information, MRI, genetics, and cognitive test. The model achieved 0.923 AUROC score for diagnosis and 0.08 MSE for prediction of antidepressant response. In addition, the results of the proposed model were quantitatively analyzed, and it confirmed that accurate diagnosis and drug response prediction are possible when the patient’s data is added. Qualitative analysis was also conducted to provide new hypotheses as well as findings on the main factors causing depression.
Route Recommendation based on Dynamic User Preference on Road Networks
http://doi.org/10.5626/JOK.2019.46.1.77
The current location based services provide maps and nearby information, or provide a route to a specific destination. A route recommendation system recommends the best route that suits the evaluation criteria for each user. The existing personalized path recommendation system recommends the route under the assumption that the user’s preference is constant regardless of the change of the time zone. However, there is a problem in that it does not reflect requirements that important factors to users can be different for each time zone, such as importance of moving distance in morning time and importance of risk in late time. In this paper, we propose a Dijkstra algorithm considering time attributes to overcome this limitation. In addition, we suggest an efficient algorithm that can search the path reflecting the change of the weight of the preference factor according to the time zone using the G-tree index structure that effectively expresses the road network.
An Efficient Method of Processing Spatio-Temporal Joins in IoT (Internet of Things) Environments
Ki Yong Lee, Minji Seo, Ryong Lee, Minwoo Park, Sang-Hwan Lee
http://doi.org/10.5626/JOK.2019.46.1.86
A spatio-temporal join is an operation that connects multiple sets of data with the same spatial and temporal values. Especially with the increasing spread of IoT (Internet of Things), the need for spatio-temporal joins is also increasing in order to retrieve data generated at the same time and location among the data generated by different things in the past. In this study, we propose an efficient method for processing spatio-temporal joins on IoT data. The proposed method divides the 3D spatio-temporal space into small subspaces and maintains the identifiers of things whose data are present in each subspace. When a spatio-temporal join between things is requested, the proposed method identifies the spaces in which the things’ data are close to each other. Then, it retrieves data contained in the identified spaces and performs the join only between them. Therefore, because only that data with the possibility of being joined are accessed, the execution cost is greatly reduced. The experimental results on a real IoT dataset show that the proposed method significantly reduces the execution time compared to the existing spatio-temporal methods.
Weather Ontology System in IoT Middleware
Yujin Kim, Soobin Jeon, Inbum Jung
http://doi.org/10.5626/JOK.2019.46.1.97
With the growing importance of weather information, the number of weather information application systems has been increased. Unfortunately, the weather application systems that currently exist neither efficiently store nor manage the vast amount of weather data obtained from various weather sensors. Additionally, they do not utilize the properties contained in the weather data making it difficult to perform intelligent searches using the semantic information present in the weather data. In this paper as a solution to the challenges mentioned, an ontology system for weather information management is constructed using IoT middleware MinT. Based on weather ontology, it is possible to efficiently and easily manage large amounts of weather sensing data by applying ontology in the Internet of Things middleware. Moreover, since inference engine and rule-base information are used, semantic properties are applied to the sensing data collected. The implemented weather ontology uses sensing data and easily provides search results to users through a UI. The usability to search results is selected as the metric for performance evaluation. In the experiments, the proposed weather ontology system exhibited high usability to search results.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr