• Journal of the Semiconductor & Display Technology
• /
• v.22 no.2
• /
• pp.128-132
• /
• 2023

Recently, with the development of deep learning technology, research on pedestrian attribute recognition technology using deep neural networks has been actively conducted. Existing pedestrian attribute recognition techniques can be obtained in such a way as global-based, regional-area-based, visual attention-based, sequential prediction-based, and newly designed loss function-based, depending on how pedestrian attributes are detected. It is known that the performance of these pedestrian attribute recognition technologies varies greatly depending on the type of backbone network that constitutes the deep neural networks model. Therefore, in this paper, several backbone networks are applied to the baseline pedestrian attribute recognition model and the performance changes of the model are analyzed. In this paper, the analysis is conducted using Resnet34, Resnet50, Resnet101, Swin-tiny, and Swinv2-tiny, which are representative backbone networks used in the fields of image classification, object detection, etc. Furthermore, this paper analyzes the change in time complexity when inferencing each backbone network using a CPU and a GPU.

• Structural Monitoring and Maintenance
• /
• v.5 no.1
• /
• pp.67-78
• /
• 2018

The stability of cutting slope influences the safety of railway operation, and how to identify the stability of the slope quickly and determine the rational monitoring plan is a pressing problem at present. In this study, the attribute recognition model of risk assessment for high cutting slope stability in the heavy haul railway is established based on attribute mathematics theory, followed by the consequent monitoring scheme design. Firstly, based on comprehensive analysis on the risk factors of heavy haul railway loess slope, collapsibility, tectonic feature, slope shape, rainfall, vegetation conditions, train speed are selected as the indexes of the risk assessment, and the grading criteria of each index is established. Meanwhile, the weights of the assessment indexes are determined by AHP judgment matrix. Secondly, The attribute measurement functions are given to compute attribute measurement of single index and synthetic attribute, and the attribute recognition model was used to assess the risk of a typical heavy haul railway loess slope, Finally, according to the risk assessment results, the monitoring content and method of this loess slope were determined to avoid geological disasters and ensure the security of the railway infrastructure. This attribute identification- risk assessment- monitoring design mode could provide an effective way for the risk assessment and control of heavy haul railway in the loess plateau.

• The Journal of Industrial Distribution & Business
• /
• v.9 no.2
• /
• pp.39-45
• /
• 2018

Purpose - The purpose of this study is to examine the image similarity and attribute recognition of the top 10 rated spa destinations (Chungnam Deoksan, Chungnam Dogo, Busan Dongrae, Daejeon Yuseong, Chungnam Asan, Gyeongbuk Bomun, Chungbuk Suanbo, Gyeongnam Jangyu, Chungnam Onyang, & Gyeongbol Bugok) in Korea based on the visits to these spa places by the customers. Research design, data, and methodology - The survey of this study was conducted on the visitors to the top 10 spa destinations in Korea from April 8 ~ April 21, 2017, and a total of 300 questionnaires were distributed. Of them, effective questionnaires used in the final study were a total of 241. In this study, empirical analysis was made through frequency analysis, factor analysis, and multidimensional scaling ALSCAL(spinning symmetry for image similarity and rectangle for attributes recognition) by using the Statistics Package SPSS 24.0. Results - According to the analysis result of spa destination image similarity, the stress level was 0.16453 and the level of the stress was good. Moreover, the coefficient of determination (RSQ) was, which had a description of each aspect of the spa destination, 0.79908. According to the results of attribute recognition, the stress value of 0.11805 represents a degree of conformity, and the coefficient of determination(RSQ) appeared at 0.98665. Therefore, the results of this analysis are that the similarities between spa destinations and the attribute recognition of the spa destinations is a suitable model that is properly expressed in two dimensions. Conclusions - First, according to the analysis result of image similarity, Deoksan & Dogo spa revealed similar images, as well as the Dongrae and Yuseong spa, while on the contrary Asan, Bomun, Suanbo spa has different images from the rest. Second, according to the results of attribute recognition, Asan and Onyang spa has competitiveness in terms of accessibility to spa destination; Yuseong, Dongrae, Jangyu spa in terms of spa facilities, spa tourism conditions, and service & shopping conditions. while spa water quality and spa costs showed low attribute reflection for all 10 spas. Therefore, the spa visitors cannot recognize the differentiation of spa water quality and spa costs.

• Journal of Electrical Engineering and Technology
• /
• v.10 no.3
• /
• pp.1264-1274
• /
• 2015

Human action recognition received many interest in the computer vision community. Most of the existing methods focus on either construct robust descriptor from the temporal domain, or computational method to exploit the discriminative power of the descriptor. In this paper we explore the idea of using local action attributes to form an action descriptor, where an action is no longer characterized with the motion changes in the temporal domain but the local semantic description of the action. We propose an novel framework where introduces local action attributes to represent an action for the final human action categorization. The local action attributes are defined for each body part which are independent from the global action. The resulting attribute descriptor is used to jointly model human action to achieve robust performance. In addition, we conduct some study on the impact of using body local and global low-level feature for the aforementioned attributes. Experiments on the KTH dataset and the MV-TJU dataset show that our local action attribute based descriptor improve action recognition performance.

• Fashion & Textile Research Journal
• /
• v.21 no.3
• /
• pp.288-299
• /
• 2019

This study examined trends in model perceptions in the Asian fashion industry through a survey on the current status of using models, model attributes, and image recognition for companies and brands participating in the Seoul Fashion Week and Hong Kong Fashion Week. The results of the study are as follows. First, an examination of the races of models used for public relations by clothing and accessory companies indicated that the use of Asian and black models was lower than white models. Second, intimacy, reliability, similarity, and professionalism were derived as attributes for a public relation model. Among these factors, only 'intimacy' showed a difference between the countries. Third, Seoul Fashion Week participants gave the highest marks for the strong individuality of the models used for their brands; however, participants in the Hong Kong Fashion Week most appreciated suitability with products and professional appearance. Fourth, the different trends of model image recognition were shown through various analysis results by country or race, in which Seoul Fashion Week participants highly perceived the global and luxurious image of white models, and were generally highly satisfied with the models. In terms of the Hong Kong Fashion Week, Asian models tended to be perceived as a more casual image, and the participants held contributions to brand recognition as the most significant factor when using Asian models.

• Journal of Information Processing Systems
• /
• v.15 no.5
• /
• pp.1171-1178
• /
• 2019

This study proposes a deep neural network model based on an encoder-decoder structure for visual dialogs. Ongoing linguistic understanding of the dialog history and context is important to generate correct answers to questions in visual dialogs followed by questions and answers regarding images. Nevertheless, in many cases, a visual understanding that can identify scenes or object attributes contained in images is beneficial. Hence, in the proposed model, by employing a separate person detector and an attribute recognizer in addition to visual features extracted from the entire input image at the encoding stage using a convolutional neural network, we emphasize attributes, such as gender, age, and dress concept of the people in the corresponding image and use them to generate answers. The results of the experiments conducted using VisDial v0.9, a large benchmark dataset, confirmed that the proposed model performed well.

• Speech Sciences
• /
• v.10 no.1
• /
• pp.71-84
• /
• 2003

In the acoustic modeling for large vocabulary speech recognition, a sparse data problem caused by a huge number of context-dependent (CD) models usually leads the estimated models to being unreliable. In this paper, we develop a new clustering method based on the C45 decision-tree learning algorithm that effectively encapsulates the CD modeling. The proposed scheme essentially constructs a supervised decision rule and applies over the pre-clustered triphones using the C45 algorithm, which is known to effectively search through the attributes of the training instances and extract the attribute that best separates the given examples. In particular, the data driven method is used as a clustering algorithm while its result is used as the learning target of the C45 algorithm. This scheme has been shown to be effective particularly over the database of low unknown-context ratio in terms of recognition performance. For speaker-independent, task-independent continuous speech recognition task, the proposed method reduced the percent accuracy WER by 3.93% compared to the existing rule-based methods.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.33 no.3
• /
• pp.477-488
• /
• 2009

As the Korean cultural contents, such as drama, films, music, gained popularity in China, Korean fashion brands used Korean celebrities as their models to as a sales promotion strategy for Chinese consumers. With the point of view that the advertising model as a human capital as well, the purpose of this study is to investigate the factors of attributes of Korean celebrity advertising model, and to analyze effects on fashion brand recognition, preference, trust and purchase intention. With convenience sampling, the research surveyed Shanghai consumers in their 20's to early 30's who had purchased Korean fashion items. The 291 responses were analyzed by frequency analysis, reliability test, factor analysis, multiple regression analysis, The results are as follows. Frist, Korean celebrity advertising model attribute factors were divided into similarity, familiarity, popularity, attractiveness and trust. Second, the brand recognition was affected by similarity, familiarity and popularity factors, and the brand preference was affected by similarity, familiarity, popularity and attractiveness factors. Third, the trust of Korean fashion brands was affected by similarity, familiarity, attractiveness, trust, brand recognition and brand preference. Lastly, the intention to purchase Korean Fashion brand was affected by similarity, familiarity, attractiveness, brand recognition, brand preference and brand trust.

• Journal of Electrical Engineering and information Science
• /
• v.2 no.4
• /
• pp.59-65
• /
• 1997

In this paper, a general fuzzy syntactic method for recognition of flaw patterns and for the measurement of flaw characteristic parameters for a non-destructive inspections signal, called eddy-current, is presented. Solutions are given to the subtasks of primitive pattern selection, signal to symbol transformation, pattern grammar formulation, and event-synchronous flaw pattern extraction based on the grammars. Fuzzy attribute grammars are used as the model for the pattern grammar because of their descriptive power in the face of uncertain constraints caused by nose or distortion in the signal waveform, due to their ability to handle syntactic as well as semantic information. This approach has been implemented and the performance of eh resultant system has been evaluated using a library of law patterns obtained from steam generator tubes in nuclear power plants by an eddy current-based non-destructive inspection method.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1996.04a
• /
• pp.670-674
• /
• 1996

In automated search of a design database to support mechanical design, it is necessaryto recognize a shape pattern which represents a design object. This paper introduces the concept of a surface relation graph (SRG) for recognizing shape patterns from a 3D boundary representation scheme of a solid model(a B-rep model). In SRG, the nodes and arcs correspond to the faces and edges shared by two adjacent faces, respectively. An attribute assigned to an arc is given by an integer which discriminates the relationship between two adjacent faces. The + sign of the integer represents the geometric convexity of the solid, and the -sign the concivity at the shared edge. The input shape is recognized by comparison with the predefined features which are subgraphs of the SRG. A hierarchyof the database for upporting the design is presented. A search for the design database is also discussed. The usefulness of this method is illustrated by some application results.