• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2022년도 추계학술발표대회
• /
• pp.308-310
• /
• 2022

With the rise of the Al wave, task-oriented dialogue systems have become one of the popular research directions in academia and industry. Currently, task-oriented dialogue systems mainly adopt pipelined form, which mainly includes natural language understanding, dialogue state decision making, dialogue state tracking and natural language generation. However, pipelining is prone to error propagation, so many task-oriented dialogue systems in the market are only for single-round dialogues. Usually single- domain dialogues have relatively accurate semantic understanding, while they tend to perform poorly on multi-domain, multi-round dialogue datasets. To solve these issues, we developed a paddle-based multi-domain task-oriented Chinese dialogue system. It is based on NEZHA-base pre-training model and CrossWOZ dataset, and uses intention recognition module, dichotomous slot recognition module and NER recognition module to do DST and generate replies based on rules. Experiments show that the dialogue system not only makes good use of the context, but also effectively addresses long-term dependencies. In our approach, the DST of dialogue tracking state is improved, and our DST can identify multiple slotted key-value pairs involved in the discourse, which eliminates the need for manual tagging and thus greatly saves manpower.

• 산업경영시스템학회지
• /
• 제46권4호
• /
• pp.101-106
• /
• 2023

With the recent surge in YouTube usage, there has been a proliferation of user-generated videos where individuals evaluate cosmetics. Consequently, many companies are increasingly utilizing evaluation videos for their product marketing and market research. However, a notable drawback is the manual classification of these product review videos incurring significant costs and time. Therefore, this paper proposes a deep learning-based cosmetics search algorithm to automate this task. The algorithm consists of two networks: One for detecting candidates in images using shape features such as circles, rectangles, etc and Another for filtering and categorizing these candidates. The reason for choosing a Two-Stage architecture over One-Stage is that, in videos containing background scenes, it is more robust to first detect cosmetic candidates before classifying them as specific objects. Although Two-Stage structures are generally known to outperform One-Stage structures in terms of model architecture, this study opts for Two-Stage to address issues related to the acquisition of training and validation data that arise when using One-Stage. Acquiring data for the algorithm that detects cosmetic candidates based on shape and the algorithm that classifies candidates into specific objects is cost-effective, ensuring the overall robustness of the algorithm.

• 한국산업융합학회 논문집
• /
• 제27권3호
• /
• pp.535-546
• /
• 2024

This research aims to develop a model to objectively and quantitatively measure the skill level of aircraft body assembly workers. Because aircraft body assembly is predominantly a manual process, skills management is a key factor of manufacturing competitiveness. Currently, skills management relies on the subjective judgment of supervisors, which lacks objectivity and reliability. As a remedy, this study proposed a systematic skill management model based on objective and quantitative evaluation criteria. By considering prior research, we developed an evaluation model that takes into account both expertise and versatility of a worker. The model selected five major tasks required for aircraft body assembly and established evaluation criteria considering the difficulty and maturity of each task. We then conducted a pilot evaluation with over 200 workers in four SMEs to validate the practicality and effectiveness of the model. Consequently, we identified and addressed the limitations of the existing evaluation method, subdivided the skill levels based on the performance capabilities of each task, and proposed a career growth path. The developed evaluation model offers critical data for executives and managers to determine work assignments, education, training, performance incentives, and wages. It is expected to enhance the attraction of new talent and systematize skills management in aviation manufacturing in the future.

• 대한임베디드공학회논문지
• /
• 제19권2호
• /
• pp.83-90
• /
• 2024

CNC tools contribute to the production of high-precision and consistent results. However, employing damaged CNC tools or utilizing compromised numerical control can lead to significant issues, including equipment damage, overheating, and system-wide errors. Typically, the assessment of external damage to CNC tools involves capturing a single viewpoint through a camera to evaluate tool wear. This study aims to enhance existing methods by using only a single manually focused Microscope camera to enable comprehensive external analysis from multiple perspectives. Applying the NeRF (Neural Radiance Fields) algorithm to images captured with a single manual focus microscope camera, we construct a 3D rendering system. Through this system, it is possible to generate scenes of areas that cannot be captured even with a fixed camera setup, thereby assisting in the analysis of exterior features. However, the NeRF model requires considerable training time, ranging from several hours to over two days. To overcome these limitations of NeRF, various subsequent models have been developed. Therefore, this study aims to compare and apply the performance of Instant NGP, Mip-NeRF, and DS-NeRF, which have garnered attention following NeRF.

• 대한안전경영과학회지
• /
• 제26권1호
• /
• pp.45-53
• /
• 2024

A Smart factories are systems that enable quick response to customer demands, reduce defect rates, and maximize productivity. They have evolved from manual labor-intensive processes to automation and now to cyber-physical systems with the help of information and communication technology. However, many small and medium-sized enterprises (SMEs) are still unable to implement even the initial stages of smart factories due to various environmental and economic constraints. Additionally, there is a lack of awareness and understanding of the concept of smart factories. To address this issue, the Cooperation-based Smart Factory Construction Support Project was launched. This project is a differentiated support project that provides customized programs based on the size and level of the company. Research has been conducted to analyze the impact of this project on participating and non-participating companies. The study aims to determine the effectiveness of the support policy and suggest efficient measures for improvement. Furthermore, the research aims to provide direction for future support projects to enhance the manufacturing competitiveness of SMEs. Ultimately, the goal is to improve the overall manufacturing industry and drive innovation.

• 한국의상디자인학회지
• /
• 제4권1호
• /
• pp.111-130
• /
• 2002

This study performed a structural questionnaire survey and non-structural interview of the ladies apparel manufacturers in Daegu on the qualification for the employees, skills required for job performance, job training, automatic manufacturing systems, and the use of computer. The results are as follows. 1. Almost all of the apparel manufacturing systems were Pair System, except one Line System in one company. In terms of outsourcing, most of the manufacturers answered “yes,” and in 1998 the outsourcing process was sewing, but in the year 2002, outsourcing has been increased :12 manufacturers(57.1%) outsourcing most of the processes except patterning, 3(14.3%) outsourcing the finish of sewing. 2. The workforce of 1998 and that of 2002 shows a significant difference(P＜. 01) between office work and management. The number of office workers has decreased from 15 down to 5.3 people. On the other hand, that of the management has slightly increased from 5.3 to 9.2 people. The number of the manual workers has decreased from 32.2 to 28.7 people. And the number of tailoring and patterning workers has slightly decreased, but the number has increased in sewing from 3.7 to 7.0 people. 3. The wage of an employee shows a significant difference between a sewing assistant(P＜. 01) and a production manager(P＜. 05), and the wage of a sewing assistant, in particular, has slightly raised from ￦905,000 to ￦1,054,000. 4. The qualifications required of employees are “cooperative human relations”(30.8％), “diligence,” and “ability for job analysis”(26.9％), and “positive thinking” (15.4％) in 1998, and “ability for job analysis”(38.5％), “cooperative human relations”(34.6％), and “positive thinking” (15.4％) in 2002. The areas for job openings are significantly different(P＜. 01) depending on the year. Job openings in the design section has increased from 1(3.8％) to 16 manufacturers (61.5％), and decreased in tailoring section from 22(84.6％) to 2 manufacturers(7.7％). Job openings in the sewing section have increased form 2(7.7％) to 6 manufacturers (23.1％). In terms of sex of the employees, there is a significant difference(P＜. 001). 19 companies(73.1％) wanted “male” in 1998, but 8 companies(30.8％) answered that they want “female” and 17 companies(65.4％) answered that “it does not matter.” About the educational background, there was a significant difference between the years. The number of the companies that want junior college graduates with an associate degree has increased(15 companies(57.7％). There was a significant difference(P＜. 05) in major of the employee. The number of the companies that want fashion majors has increased from 5(19.2％) to 20(76.9％). 5. In terms of job skills required, there was no significant difference. In 1998, “production skills” (46.2％) and “ability for job analysis” (26.9％) were required, and in 2002, “ability for job analysis” (42.3％) and “emotional skills” (26.9％). 6. In regard to training for job skills, “fashion professional training” has slightly decreased from 65.4％ in 1998 to 46.2％ in 2002, however, “training for job analysis” has slightly increased from 30.8％ in 1998 to 46.2％ in 2002, which indicates the fact that “fashion professional training” and “ability for job analysis” have been emphasized. 7. The number of the manufacturers purchased apparel CAD has increased from 1(3.8％) to 3(11.5％), and the number of the manufacturers that have no plan for purchase has increased from 16(61.5％) in 1998 to 15(57.7％), still taking up a big proportion. 8. About the use of computers in manufacturing, there is a significant difference(P＜. 05). The number of the manufacturers using computer has increased from 5(19.2％) to 15(57.7％) and that of the manufacturers which do not use computers has decreased from 17(57.7％) to 8(30.8％). 9. In the interviews with the owners of the manufacturers, they pointed that schools should give more weight on practical training courses, the invitation of experts in the specific field, complex production systems, training courses for sewing, field trip courses, and furthering specialty education, personality and vocational education.

• 한국응급구조학회지
• /
• 제14권3호
• /
• pp.13-28
• /
• 2010

Purpose: In this study, we investigated the better application of the law which is about the AED installation and more effective ways of emergency medical care system, to understand the law and to research the current condition of public facilities which belong to local governments, and to seize the aspect of safety guards who currently work in order to provide the installation of AED in the public facilities and to provide more efficient emergency medical service with the effectuation of the immunity law of the good intention of first-aid treatment. Methods: In Gwang-ju, 234 public facilities have been identified by 31 December, 2008. With the exception of the duplication, we researched 158 facilities and received the answers from 95 of them. Results: In the research, 53% of them have had internal emergency first-aid education, and 55% of them didn't have this education and a CPR education manual, and 30% of the facilities even didn't know how to connect with the manager of the company for the first-aid department. On the other hand, most of them were highly interested in CPR and AED education on the ratio of 91% and 93%. 88% of them have been trained about first-aid, 51% of them haven't been retrained, 17% have never been trained. so, the reality of emergency system at public facilities is serious. 78% of them knew they are working at public facilities, though 49% of them didn't know about AED installation. 57% of them didn't know the fact there is the immunity law related with good intentions for first-aid treatment. 63% of the facilities have security guards, and 30% of them didn't answer the questions. Also, many of them agreed to the opinion that all employees should have first-aid training. At representative survey report of participator of public-facility, emergency treatment is 61%, 16% of patients calling. Accordingly they importantly think better doing an on-site first-aid than evacuating the patient. And the rates show that 57% of them answerers tend to call Fire-Office(119) for evacuating the patients, and 28% of them EMIC(1339) for the first-aid. Conclusions: In this study, we are suggest to improve the details of the efficient operations and management after the grasp of the uninstallation, indifference, and unreliable conditions of AED. 1) Need a publicity of AED install cognition which is an emergency medical instrument at public facilities. 2) Arrangement of safety agents at facilities and concerns about them for good management from the parties concerned. 3) Need a designation of legal details according to the decision of the AED installation and the standard of the AED installation. 4) Training about first-aid of safety guards and the persons concerned in the facilities should be practiced participation with the positive and through this, first-aid treatment could be done by anyone who knows the immunity law related to medical emergency. 5) The brochures for the potential users and the results form practicing the instructions need to be improved in many ways through recording the emergency cases that have happened.

• 한국지구과학회지
• /
• 제28권1호
• /
• pp.14-23
• /
• 2007

본 연구의 목적은 초임 지구과학 교사들이 야외 지질 학습에 대하여 어떻게 인식하고 있는지를 알아보는 것이다. 이를 위하여 2006년 1급 정교사 자격 연수를 받고 있는 지구과학 교사 116명을 대상으로 야외 지질 학습의 필요성과 교육적 가치, 교사의 야외 지질 학습 경험, 야외 지질 학습의 어려운 점과 바람직한 야외 지질 학습의 방법에 관한 인식을 조사하였다. 연구 결과, 초임 지구과학 교사들은 지구과학에서 야외 지질 학습은 필요하다고 인식하고 있었고, 교육적 가치에 대해서도 긍정적인 반응을 보였다. 대부분의 교사들은 교사양성과정에서 야외 지질 답사를 경험하였지만, 야외 지질 학습을 지도하는 방법을 배운 교사는 많지 않았다. 교직에 들어와서 야외 지질 답사를 다녀본 경험이나 학생들을 인솔하여 야외 지질 학습을 지도한 경험은 적은 것으로 나타났다. 많은 지구과학 교사들이 야외 지질 학습을 지도하기 어렵다고 인식하였으며, 가장 어려운 이유로 교사 자신의 경험 부족이라고 인식하였다. 이를 해결하기 위한 방안으로 현직 교사 연수를 통한 야외 지질 학습의 교육이 필요하다고 제안하였다 또한 바람직한 야외 지질 학습이 이루어지기 위해서는 야외 지질 학습의 사전 활동으로 야외 지질 학습 장소의 지질학적인 소개나 설명이 필요하다고 보았고, 활동 중에는 교사의 안내와 학생들의 협동학습이 바람직하다고 인식하였다. 그리고 사후 활동으로는 야외 지질 학습 결과에 대한 발표 및 토론이 적합하다고 보았다.

• 방송공학회논문지
• /
• 제24권6호
• /
• pp.1076-1084
• /
• 2019

본 논문에서는 딥러닝을 적용하여 미디어 내의 등장인물 및 사물을 인식, 메타데이터를 추출하고 이를 통해 아카이브를 구축하는 시스템을 개발하였다. 방송 분야에서 비디오, 오디오, 이미지, 텍스트 등의 멀티미디어 자료들을 디지털 컨텐츠로 전환하기 시작한지는 오래 되었지만, 아직 구축해야 할 자료들은 방대하게 남아있다. 따라서 딥러닝 기반의 메타데이터 생성 시스템을 구현하여 미디어 아카이브 구축에 소모되는 시간과 비용을 절약 할 수 있도록 하였다. 전체 시스템은 학습용 데이터 생성 모듈, 사물 인식 모듈, 등장인물 인식 모듈, API 서버의 네 가지 요소로 구성되어 있다. 미디어 내에서 등장인물 및 사물을 인식하여 메타데이터로 추출할 수 있도록 딥러닝 기술로 사물 인식 모듈, 얼굴 인식 모듈을 구현하였다. 딥러닝 신경망을 학습시키기 위한 데이터를 구축하기 용이하도록 학습용 데이터 생성 모듈을 별도로 설계하였으며 얼굴 인식, 사물 인식의 기능은 API 서버 형태로 구성하였다. 1500명의 인물, 80종의 사물 데이터를 사용하여 신경망을 학습시켰으며 등장인물 테스트 데이터에서 98%, 사물 데이터에서 42%의 정확도를 확인하였다.

• Archives of Plastic Surgery
• /
• 제37권4호
• /
• pp.391-395
• /
• 2010

Purpose: Given that the critical nature of the microvascular anastomosis to what is often a long and difficult reconstructive operation, trainees need to have a high level of microsurgical competence before being allowed to perform microsurgery on patients. Some artificial substitutes and dead or live animal models have been used to improve manual dexterity under the operating microscope. Yet, most surgeons are not equipped with such models, so search for easy available and appropriate microsurgical practice model have been an issue. Umbilical artery, placental vessels and gastroepiploic arteries have been previously suggested as a microsurgical training model, which involves other surgical departments. The purpose of this article is to introduce that saphenous vein specimen obtained from varicose vein surgery is useful and has many advantages as training model for the practice of microvascular anastomosis. Methods: The conventional technique using perforation/inversion method with a metallic stripper is widely performed for varicose vein patients. The stripper is inserted through disconnected safeno-femoral junction and retrieved at the knee or the medial side of ankle. The length of saphenous vein specimens removed is about that of one's leg and inversed from inside out. Obtained saphenous vein specimens are re-inversed and cleansed with normal saline, to be readily available for microsurgical practice. Preserved in a squeezed wet saline gauze and refrigerated, frozen or glycerated specimens were investigated into their comparative quality for microsurgical practice. Results: Varicose vein surgery remains one of the common operations performed in the field of plastic surgery. Convenient informed consent regarding the vessel donation can be easily signed. The diameter of the obtained saphenous vein is as variable as 1.5 to 6 mm, which is already stripped, and is in sufficient length corresponding to that of patient's leg. Vessels specimens were available for microsurgical practice within 1 week period when preserved with squeezed wet saline gauze, and the preservation period could be extended monthly by freezing it. Conclusion: Saphenous vein obtained from varicose vein patients provide with variable size of vessel lumen with sufficient length. The practice can be cost effective and does not require microsurgical laboratory. Additionally there is no need of involving other surgical departments in acquiring vessel specimens. Furthermore, simple preservation method of refrigerating for a week or freezing with squeezed wet saline gauze for a month period, allow the saphenous vein obtained after varicose vein surgery as an excellent model for the microsurgical practice.