• Journal of Convergence for Information Technology
• /
• v.7 no.2
• /
• pp.31-36
• /
• 2017

In this paper, we try to search and reproduce the image part of a specific cast from a large number of images. The conventional method must manually set the offset value when searching for a scene or viewing a corner. However, in this paper, the proposed method learns the main character 's face, then finds the main character in the image recognition and moves to the scene where the main character appears to reproduce the image. Data for specific performers is extracted and collected using crawl techniques. Based on the collected data, we learn using convolutional neural network algorithm and perform performance evaluation using it. The performance evaluation measures the accuracy by extracting and judging a specific performer learned in the extracted key frame while playing the drama. The performance confirmation of how quickly and accurately the learned scene is searched has obtained about 93% accuracy. Based on the derived performance, it is applied to the image service such as viewing, searching for person and detailed information retrieval per corner

• Asia Marketing Journal
• /
• v.22 no.4
• /
• pp.27-45
• /
• 2021

This study examines the effects of celebrity endorsed advertising from a new perspective of the prior research that emphasizes the matchup between the brand and the celebrity. Due to the recent sharing experiences of the celebrity and their fans on SNS, it is hypothesized that the shared stories would impact viewers' responses that are often expressed in their likes, dislikes, shares and comments on SNS. In this study, the episodic type of advertising is hypothesized to have more favorable and active responses from viewers than the typical celebrity image-focused ads would have. By crawling and analyzing viewers' responses on YouTube toward 12 BTS endorsed ads, the hypotheses are confirmed as higher ratio of likes, lower ratio of dislikes and significantly higher ratio of comments over both total views and total likes were found. For the rationale behind, total 1800 comments were categorized into 4 major content types such as attached, experiential, empathic and self-related ones that are all considered as important factors influencing the strong ad effect. The results showed that the episodic ads have marginally more emotional comments than the celeb image ads. The difference was only found in experiential and empathic responses but not in self-related responses. Contrary to the hypothesis, the comments expressing attachment were found more for the celebrity image-focused ads than the episodic ones. It does not seem to suggest that the celebrity image focused ads are better to capture viewers' attachment towards the celebrity and the ad endorsed, but that the episodic ads draw viewers into relatively deeper level of attachment such as empathy by perceiving the authenticity of the celebrity and the brand. In conclusion, the shared stories on SNS can be a factor in the match-up theory on celebrity endorsed ad effects.

• KIISE Transactions on Computing Practices
• /
• v.24 no.1
• /
• pp.53-57
• /
• 2018

A poisoning accident is often caused by a situation in which people eat poisonous mushrooms because they cannot distinguish between edible mushrooms and poisonous mushrooms. In this paper, we propose an automatic mushroom recognition system by using the convolutional neural network. We collected 1478 mushroom images of 38 species using image crawling, and used the dataset for learning the convolutional neural network. A comparison experiment using AlexNet, VGGNet, and GoogLeNet was performed using the collected datasets, and a comparison experiment using a class number expansion and a fine-tuning technique for transfer learning were performed. As a result of our experiment, we achieve 82.63% top-1 accuracy and 96.84% top-5 accuracy on test set of our dataset.

• Fashion & Textile Research Journal
• /
• v.17 no.5
• /
• pp.731-740
• /
• 2015

Mobile (smartphone) search engine marketing is increasingly important. Accordingly, the development of visual apparel search technology to obtain easier and faster access to visual information in the apparel field is urgently needed. This study helps establish a proper classifying system for an apparel search after an analysis of search techniques for apparel search applications and existing domestic and overseas apparel sites. An annotation technique is developed in accordance with visual attributes and apparel categories based on collected data obtained by web crawling and apparel images collecting. The categorical composition of apparel is divided into wearing, image and style. The web evaluation site traces the correlations of the apparel category and apparel factors as dependent upon visual attributes. An appraisal team of 10 individuals evaluated 2860 pieces of merchandise images. Data analysis consisted of correlations between apparel, sleeve length and apparel category (based on an average analysis), and correlation between fastener and apparel category (based on an average analysis). The study results can be considered as an epoch-making mobile apparel search system that can contribute to enhancing consumer convenience since it enables an effective search of type, price, distributor, and apparel image by a mobile photographing of the wearing state.

• The Research Journal of the Costume Culture
• /
• v.29 no.4
• /
• pp.603-615
• /
• 2021

The aim of this study was to analyze the semantic network structure of keywords and the visual composition of images extracted from Instagram in relation to the multi-persona phenomenon with in fashion imagery, which has recently been attracting attention. To this end, the concept of a 'secondary character', which forms a separate identity from a 'main character' on various social media platforms as well as on the airwaves, was considered as the spread of multi-persona and #SecondaryCharacter on Instagram was investigated. 3,801 keywords were collected after crawling the data using Python and morphological analysis was undertaken using KoNLP. The semantic network structure was then examined by conducting a CONCOR analysis using UCINET and Netdraw to determine the top 50 keywords. The results were then classified into a total of 6 clusters. In accordance with the meaning and context of the keywords included in each cluster, group names were assigned : virtual characters, relationship with the main character, hobbies, daily record, N-job person, media and marketing. Image analysis considered the technical, compositional, and social styles of the media based on Gillian Rose's visual analysis method. The results determined that Instagram uses fashion images that virtualize one's face to produce multi-persona representation s that show various occupations, describe different types of hobbies, and depict situations pertaining to various social roles.

• Korean Journal of Artificial Intelligence
• /
• v.11 no.1
• /
• pp.25-29
• /
• 2023

In this paper, the purpose is to create a standard of AI training dataset type for commercial space design. As the market size of the field of space design continues to increase and the time spent increases indoors after COVID-19, interest in space is expanding throughout society. In addition, more and more consumers are getting used to the digital environment. Therefore, If you identify trends and preemptively propose the atmosphere and specifications that customers require quickly and easily, you can increase customer trust and conduct effective sales. As for the data set type, commercial districts were divided into a total of 8 categories, and images that could be processed were derived by refining 4,009,30MB JPG format images collected through web crawling. Then, by performing bounding and labeling operations, we developed a 'Dataset for AI Training' of 3,356 commercial space image data in CSV format with a size of 2.08MB. Through this study, elements of spatial images such as place type, space classification, and furniture can be extracted and used when developing AI algorithms, and it is expected that images requested by clients can be easily and quickly collected through spatial image input information.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2018.07a
• /
• pp.463-465
• /
• 2018

공유저작물을 이용하는데 있어서 권리관리정보(RMI, Rights Management information)를 제대로 파악하지 못하거나 제대로 된 정보를 제공 받지 못하는 경우 이용자들은 저작권분쟁에 휘말릴 수 있다. 이러한 이유로 공유저작물을 제공하는 사이트에서는 공유저작물에 대한 정확하고 최신의 RMI 정보를 제공하기 위하여 통합하여 관리하고 최신의 정보로 갱신해야 한다. 하지만 동일한 권리를 가진 이미지는 다양한 이미지 포맷과 사이즈 변경에 따라 다른 형태로 유통되기 때문에 이에 대한 갱신처리가 중요하다. 본 논문에서는 이미지 특징점 기술을 활용하여 권리관리정보에 대한 중복데이터 문제를 해결할 수 있는 기법을 제시한다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.01a
• /
• pp.227-229
• /
• 2021

본 논문에서는 노약자 및 장애인의 증가로 인한 조행보조기구 사용량이 증가하고 있으나 물리적인 보조기구는 있지만 AI를 통한 서비스와 보조보행기구에 관한 AI 데이터셋이 부족하다. 이러한 문제점을 보안하기 위해 본 논문에서는 상기 데이터셋을 설계 및 구축하기 위해 Node JS를 사용하여 이미지 크롤링 프로그램을 구현하여 이미지 데이터를 수집했으며, Yolo Maker를 활용하여 수집된 이미지를 데이터셋으로 변환시켰다. 이를 통해 노약자 및 장애인을 위한 AI 서비스 구축에 필요한 데이터를 손쉽게 설계 및 구축한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.719-722
• /
• 2021

최근 딥러닝은 이미지 인식, 음성 인식 등 여러 분야에서 고려되고 있는 기술이다. 그러나 딥러닝 기술을 이용하기 위해서는 대형데이터 세트가 필요하나 이를 구축하기 힘들고 많은 시간이 필요하다는 문제점이 있다. 이에, 본 논문에서는 웹 크롤링을 통해 사용자가 원하는 카테고리의 이미지 데이터 세트를 수집하고 수집한 데이터들을 전처리 과정을 통해 딥러닝 모델에 입력할 수 있는 데이터 세트의 구축을 자동화하며, 전이학습을 통해서 적은 훈련 시간과 높은 정확도를 얻을 수 있는 이미지 분류모델을 제안한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.758-759
• /
• 2023

이 논문에서는 웹 크롤링과 인공지능의 색상 추출 기능을 사용하여 사용자에게 맞는 향수를 추천해주는 서비스를 구현한다. 웹 사이트 제작에 용이한 Java 와 웹 크롤링과 인공지능 구현에 용이한 Python 을 기반으로 구현하였다.