• 한국작물학회:학술대회논문집
• /
• 한국작물학회 2022년도 추계학술대회
• /
• pp.80-80
• /
• 2022

Estimation of heading date of paddy rice is laborious and time consuming. Therefore, automatic estimation of heading date of paddy rice is highly essential. In this experiment, deep learning classification models were used to classify two difference categories of rice (vegetative and reproductive stage) based on the panicle initiation of paddy field. Specifically, the dataset includes 444 slanted view images belonging to two categories and was then expanded to include 1,497 images via IMGAUG data augmentation technique. We adopt two transfer learning strategies: (First, used transferring model weights already trained on ImageNet to six classification network models: VGGNet, ResNet, DenseNet, InceptionV3, Xception and MobileNet, Second, fine-tuned some layers of the network according to our dataset). After training the CNN model, we used several evaluation metrics commonly used for classification tasks, including Accuracy, Precision, Recall, and F1-score. In addition, GradCAM was used to generate visual explanations for each image patch. Experimental results showed that the InceptionV3 is the best performing model in terms of the accuracy, average recall, precision, and F1-score. The fine-tuned InceptionV3 model achieved an overall classification accuracy of 0.95 with a high F1-score of 0.95. Our CNN model also represented the change of rice heading date under different date of transplanting. This study demonstrated that image based deep learning model can reliably be used as an automatic monitoring system to detect the heading date of rice crops using CCTV camera.

• 한국빅데이터학회지
• /
• 제5권1호
• /
• pp.189-195
• /
• 2020

본 논문은 딥러닝을 위한분산학습에서학습속도를 저하시키는 stale 문제를 최소화하기 위한 방법으로 데이터 시퀀싱을 제안하였다. 이데이터 시퀀싱 문제는일반 납기를 갖는 단일 공정 하에서 일찍 혹은 늦음 정도의 총합을 최소화 하는 스케줄링 문제로 모델링할 수 있다. 만약 최적해에서 크기가 작은 작업과 큰 작업의 순서가 미리 알려져 있다면, 이 스케줄링 문제가 효율적으로 풀린다는 것을 보였다.

• 한국항해항만학회지
• /
• 제34권4호
• /
• pp.287-292
• /
• 2010

The purpose of this study was to create a cyber-training & education program in response to the needs of skippers and crews operating tug-barges within Korean coastal waters and the rapid changers in this industry. Skippers and crews are inclined to operate tug-barges on the basis of experience rather than information. It is not easy to provide useful information whenever they want or to drill them in safety management skills, because of their passive attitude toward education and the few opportunities that exist. In order to increase educational opportunities, efficiency and motivation, the authors have developed this program which consists of a 'tug bridge resource management module, risk perception training module, accident case module, operating module and navigation module', and are hoping that this program will enhance and strengthen all tug-barge operations. We are also putting all our energies into designing up to date animation programs and developing new scenarios concerning the method of evaluation and certification distribution.

• 한국가정과교육학회지
• /
• 제13권2호
• /
• pp.85-99
• /
• 2001

The purpose of this research is intended to analyze the actual condition of in-service training for the teachers who teach Home Economics Education in the middle and high school and to survey their viewpoints on the actual condition of the in-service training. So I could find the effective ways on the in-service training by finding many kinds of needs and improvements on the basis of present problems in the Home Economics Education. The research results are as follows : 1. This research indicated that most teachers answered that the most proere significant per time of the in-service training was the school vacations(49.5%). 63.4% of the teachers answered that the present 60 hours on the question of how many hours are appropriate is proper. On the question of what the most proper cycle for the educational training is. 47.2% of the teachers answered that the present 3-year cycle is appropriate. 35% of them supported the selecting system for the trainee according to the experienced or non-experienced for the up-to-date in-service training. And 35% of them answered that the speakers for the training program must be the experienced teachers in education. In the contents of the training program. many teacher insisted that the percentage of the text for the major should be raised(49.6%). According to the survey about the teaching and learning methods and evaluation. teachers were satisfied with the levels of satisfaction on the teaching and learning methods in 65.1%. This survey indicated that teacher preferred the discussion and case study(35.9%). the practice-centered class(29.3%) and the on-the-spot study(20.9%) in order on the teaching methods. In terms of the educational environment. 56.9% of the teachers answered the number of trainee is too many and they suggested that the proper number of trainee is about 20∼30. 2. This research showed that the most important problem of this training system was the over-population of the trainee(33.5%) and the most severe problem of the educational environment was the lack of air-cooled and heated system(24.8%).

• 한국CDE학회논문집
• /
• 제15권5호
• /
• pp.383-392
• /
• 2010

The purpose of this study is to develop three Judo-doll systems for enhancement of Judo's performance. Traditional Judo training requires a human training partner. Unfortunately it is not always easy to find one. Multifunctional Judo-doll training system has therefore been developed, and is described here. The system consists of a dummy, a power generating mechanism, and kinematic links. The power-generating mechanism generates forces similar to those of a human, by adjusting deadweights and controlling powderbrake's forces. The powderbrake force is controlled by the microprocessor according to the exercise scenario. The kinetic links, which mimic a human training partner's motions, has been developed based on a $Vicon^{TM}$ system's analysis of the movement of human training partners. This mechanism whose name is "L link-wire" consists of L type links, wire, roller, and dead weight. This mechanism generates the force that leads the link to the neutral position regardless the link is pushed or pulled. The lifting mechanism that lifts the doll when the one-armed shoulder throw skill is applied is also developed. A 32-bit microprocessor controls the whole system; it reads the loadcell data, controls the electromagnetic force, and communicates with a PC via Bluetooth. The training history, including loadcell data, date, and training time, is stored in the PC for analysis. This training system can be used to enhance the Judo performance of any self training player.

• 한국작물학회지
• /
• 제27권2호
• /
• pp.141-146
• /
• 1982

Hop의 증수를 위한 재배기술의 일환으로 적심효과를 검토하고자 무적심, 1회적심, 2회적심, 3회적심의 적심방법 시험을 1979년에 실시하였고, 무적심, 5월 14일 적심(주경 3절위 신장시), 5월 19일 적심(주경 4절위 신장시), 5월 24일 적심(주경 5절위 신장시), 5월 29일 적심(주경 6절위 신장시)의 주경적심시험을 1980년에 실시하였던 바, 그 결과를 요약하면 다음과 같다. 1. 무적심에 비하여 적심회수가 거듭되거나 적심시기가 늦어질수록 영양생장기간 및 개화기가 지연되었다. 2. 주경적심처리는 무적심보다 만장과 총측지수 적었으나 측지 생육이 촉진되어 조기 적심일수록 만중이 증가되었을 뿐만 아니라, 주경 10절위 이하에서도 구화가 착생되어 착화측지수가 증가되었다. 3. 조기적심일수록 무적심에 비하여 10절위 이상 20절위 이하의 측지장이 현저히 길어져 착화측지당구화수 및 주당 구화수가 증가되어 12-43%의 증수를 가져왔다. 4. 만기적심(5월 29일 적심)은 적심 후에서 모화기까지의 일수가 짧아 무적심보다 착화측지수가 적었고, 측지장이 짧아 주당 구화수가 적었으며 100구화중도 가벼워 무적심보다 20% 감수되었다.

• The Journal of Korean Physical Therapy
• /
• 제15권3호
• /
• pp.202-222
• /
• 2003

Chronic upper extremity hemiparesis is a leading cause of functional disability after stroke. The purpose of this study were to identify effects of a 6weeks repetitive bilateral arm training on upper motor function and the reorganization of motor network. Four chronic stroke patients participated in this study. They performed for 6 consecutive weeks, 3 days a week, 30 minutes a day. In the single group study, four 5-minute periods per session of bilateral arm training were performed with the use of a custom-designed arm training machine. The results of this study was as follows. 1. Following the 6weeks period of RBAT, patient exhibited a improvement in FMA and BBT. 2. Following the 6weeks period of RBAT, it showed improvement in reaching time, symbol digit substitution and finger tapping speed of KCNT. 3. fMRI activation after RBAT showed a focal map in lesional cortical area and perilesional motor areas. These fMRI data suggest that hemodynamics response to RBAT reflect sensorimotor reorganization in contralateral hemisphere. In conclusion, these date suggest that improved upper extremity function induced by repetitive bilateral arm training after stroke is associated with reorganization of motor network as a neural basis for the improvement of paratic upper extremity function.

• 대한통합의학회지
• /
• 제1권2호
• /
• pp.23-35
• /
• 2013

Purpose : We intend to make the study date for an effect of therapy by comparing the functional level both before and after conducting handwriting training and strength training as a part of treatment to improve muscle strength and function of the patient's non-dominant hand. Method : 8 subjects in writing training group conducted hand writing training 30 minutes at once and three times a week for 4 weeks in total 12 times, and 8 subjects in muscle training group conducted muscle training program of putty and Rolyan ergonomic hand exerciser for 15 minutes respectively in sum up 30 minutes at once and three times a week for 4 weeks in total 12 times. 8 subjects in control group are not applied any training for 4 weeks. Results : It was much more effective in handwriting training than muscle strength training by Grooved pegboard because this study showed the speed decrease from 67.11 to 58.26 seconds in handwriting compared with muscle strength training which showed 5.22 seconds decrease from 67.54 to 62.32(P<.05). It showed about 1.34 muscle strength improvement from 6.60 to 7.94 in handwriting training and 0.92 improvement of muscle strength from 7.04 to 7.96 in muscle strength training by 3-jaw chuck pinch, so handwriting training was more effective(P<.05). It showed 11.58 seconds decrease in handwriting training from 26.62 to 18.01 seconds and 10.93 seconds decrease from 27.43 to 16.50 seconds in muscle strength training, so it was significantly shortened both in handwriting and muscle strength training(P<.05). Conclusion : Dexterity, muscle strength, and handwriting ability of non-dominant hand could improve both the handwriting training and the muscle strength training.

• 상하수도학회지
• /
• 제12권4호
• /
• pp.43-52
• /
• 1998

The purpose of this study is to establish a method of estimating the daily urban water demend using Backpropagation algorithm is part of ANN(Artificial Neural Network). This method will be used for the development of the efficient management and operations of the water supply facilities. The data used were the daily urban water demend, the population and weather conditions such as treperarture, precipitation, relative humidity, etc. Kwangju city was selected for the case study area. We adjusted the weights of ANN that are iterated the training data patterns. We normalized the non-stationary time series data [-1,+1] to fast converge, and choose the input patterns by statistical methods. We separated the training and checking patterns form input date patterns. The performance of ANN is compared with multiple-regression method. We discussed the representation ability the model building process and the applicability of ANN approach for the daily water demand. ANN provided the reasonable results for time series forecasting.

• KEPCO Journal on Electric Power and Energy
• /
• 제8권2호
• /
• pp.137-141
• /
• 2022

To inspect the power facility faults using artificial intelligence, it need that improve the accuracy of the diagnostic model are required. Data augmentation skill using generative adversarial network (GAN) is one of the best ways to improve deep learning performance. GAN model can create realistic-looking fake images using two competitive learning networks such as discriminator and generator. In this study, we intend to verify the effectiveness of virtual data generation technology by including the fake image of power facility generated through GAN in the deep learning training set. The GAN-based fake image was created for damage of LP insulator, and ResNet based normal and defect classification model was developed to verify the effect. Through this, we analyzed the model accuracy according to the ratio of normal and defective training data.