• Journal of Intelligence and Information Systems
• /
• v.28 no.3
• /
• pp.45-55
• /
• 2022

Recently, large-scale language models (LPLM) have been shown state-of-the-art performances in various tasks of natural language processing including intent classification. However, fine-tuning LPLM requires much computational cost for training and inference which is not appropriate for dialog system. In this paper, we propose compressed intent classification model for multi-agent in low-resource like CPU. Our method consists of two stages. First, we trained sentence encoder from LPLM then compressed it through knowledge distillation. Second, we trained agent-specific adapter for intent classification. The results of three intent classification datasets show that our method achieved 98% of the accuracy of LPLM with only 21% size of it.

• The Journal of Korean Academy of Sensory Integration
• /
• v.15 no.2
• /
• pp.46-65
• /
• 2017

Objective : The purpose of this research is to find clinical effects of application of weighted vest during task-oriented training focused on gross motor performance and balance abilities of children with spastic diplegia. Methods : 34 subjects were divided by simple random sampling into two groups; experimental group (male : 9, female : 8, average age : 8.12) and placebo group (male : 9, female : 9, average age : 7.53). Both two groups underwent to 40 minute intervention, twice a week for 12 weeks. The intervention was task-oriented training focused on facilitating closed kinematic chain and multi-joint functional movement pattern. During the training, the experimental group received loaded-resistance weighted vest and placebo group also received weighted vest but without loaded-resistance. Participants in both groups underwent 8 to 10 reps of the task-oriented training and there were 3 minutes break time between tasks. There were pre-test of gross motor performance and balance abilities, and two times of post-tests were performed upon 6 weeks and 12 weeks after the intervention completed. And in final, an additional follow-up test was performed 12 weeks after the evaluation was finished in order to find any difference between the two groups over time. Results : There was significant difference in Gross Motor Performance Measure (GMPM) between two groups. It is found that average score of the experimental group increased more than the placebo group after 6 weeks and 12 weeks intervention (p<.05). There was significant difference in Pediatric Berg's Balance Scale (PBS) between two groups. It is found that average score of the experimental group increased more than the placebo group after 6 weeks and 12 weeks intervention (p<.05). Conclusion : Based on the results in this study, it is proposed that application of weighted vest into task-oriented training to facilitating closed kinematic chain and multi-joint movement can improve gross motor performance and balance abilities of children with cerebral palsy.

• Smart Media Journal
• /
• v.12 no.9
• /
• pp.45-59
• /
• 2023

Weakly supervised object localization (WSOL) is a task of localizing an object in an image using only image-level labels. Previous studies have followed the conventional class activation mapping (CAM) pipeline. However, we reveal the current CAM approach suffers from problems which cause original CAM could not capture the complete defects features. This work utilizes a convolutional neural network (CNN) pretrained on image-level labels to generate class activation maps in a multi-scale manner to highlight discriminative regions. Additionally, a vision transformer (ViT) pretrained was treated to produce multi-head attention maps as an auxiliary detector. By integrating the CNN-based CAMs and attention maps, our approach localizes defective regions without requiring bounding box or pixel-level supervision during training. We evaluate our approach on a dataset of apple images with only image-level labels of defect categories. Experiments demonstrate our proposed method aligns with several Object Detection models performance, hold a promise for improving localization.

• The Journal of Korean Physical Therapy
• /
• v.32 no.3
• /
• pp.185-190
• /
• 2020

Purpose: Gait and cognitive impairment in stroke patients exacerbate fall risk and mobility difficulties during multi-task walking. Virtual reality can provide interesting and challenging training in a community setting. This study evaluated the effect of community-based virtual reality gait training (VRGT) using a 360-degree image on the gait ability of chronic stroke patients. Methods: Forty-five chronic stroke patients who were admitted to a rehabilitation hospital participated in this study. Patients meeting the selection criteria were randomly divided into a VRGT group (n=23) and a control group (n=22). Both these groups received general rehabilitation. The VRGT group was evaluated using a 360-degree image that was recorded for 50 minutes a day, 5 days per week for a total of 6 weeks after their training. The control group received general treadmill training for the same amount of time as that of the VRGT group. The improvement in the spatiotemporal parameters of gait was evaluated using a gait analyzer system before and after training. Results: The spatiotemporal gait parameters showed significant improvements in both groups compare with the baseline measurements (p<0.05), and the VRGT group showed more improvement than the control group (p<0.05). Conclusion: Community-based VRGT has been shown to improve the walking ability of chronic stroke patients and is expected to be used in rehabilitation of stroke patients in the future.

• 제어로봇시스템학회:학술대회논문집
• /
• 1991.10a
• /
• pp.1028-1033
• /
• 1991

This paper describes the design of a neural network estimator to estimate weld pool sizes for on-line use of quality monitoring and control in GMA welding processes. The estimator utilizes surface temperatures measured at various points on the top surface of the weldment as its input. The main task of the neural net is to realize the mapping characteristics from the point temperatures to the weld pool sizes through training, A series of bead-on plate welding experiments were performed to assess the performance of the neural estimator.

• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.5 s.98
• /
• pp.112-123
• /
• 1999

Virtual reality is becoming a powerful tool for various applications such as training, entertainment, surgery, tele-robotics etc. One potential use for virtual reality is to allow several users to interact in a single virtual environment, for example several students sitting in front of different computers connected over a network. In this paper, we present a loosely coupled architecture of haptic display in the multi-user virtual world. The method of controlling haptic devices as well as the way of configuring individual haptic display system are addressed. We will develop an experimental virtual reality system for two remote users and conclude with an experimental work for the task of a multi-player ping-pong and grasping of a common object.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09a
• /
• pp.480-483
• /
• 2003

We propose a new framework that aims at multi-purpose image recognition, a difficult task for the conventional rule-based systems. This framework is farmed based on the idea of computer-based learning algorithm. In this research, we introduce the new functions of an additional learning and a knowledge reconstruction on the Fuzzy Inference Neural Network (FINN) (1) to enable the system to accommodate new objects and enhance the accuracy as necessary. We examine the capability of the proposed framework using two examples. The first one is the capital letter recognition task from UCI machine learning repository to estimate the effectiveness of the framework itself, Even though the whole training data was not given in advance, the proposed framework operated with a small loss of accuracy by introducing functions of the additional learning and the knowledge reconstruction. The other is the scenery image recognition. We confirmed that the proposed framework could recognize images with high accuracy and accommodate new object recursively.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.11
• /
• pp.39-46
• /
• 2022

In this paper, we propose a multi-task model that can simultaneously predict general-purpose tasks such as part-of-speech tagging, lemmatization, and dependency parsing using the UD Korean Kaist v2.3 corpus. The proposed model thus applies the self-attention technique of the BERT model and the graph-based Biaffine attention technique by fine-tuning the multilingual BERT and the two Korean-specific BERTs such as KR-BERT and KoBERT. The performances of the proposed model are compared and analyzed using the multilingual version of BERT and the two Korean-specific BERT language models.

• Korean Security Journal
• /
• no.14
• /
• pp.215-234
• /
• 2007

Our country security industry 1960's service expense of the beginning U.S. army unit it accomplishes the growth which is quick with start, currently about 2,500 triumph the security enterprises which it goes over are being active. But the majority in these enterprise about lower cotton can a forever characteristic with pressure and the manpower civil official ability insufficient back of faithlessness management and capital power. To sleep with afterwords it presents the security dogs deployment plan for an efficient security together from the research which it sees hereupon and it does. First, it cultivates the domestic mountain progress dog which is a breed which is suitable with the security dogs and the shovel flesh dog back with the security dogs. Specially the Jindo of the breed which is excellent training which is suitable in task of the security dogs it leads and if it uses appropriately, it industrializes our specific the Jindo and protection there is a possibility of getting the effect which falls to also the gist which it rears rightly. It cultivate the second, security dogs and it magnifies training. The security dogs consequently is it will be able to accomplish the task above 2 branches to training method. Namely, after finishing obedience training, it is to be in security activity it will execute guard or detection back special training which is suitable in task and it will be able to commit. Third, it uses the security dogs which is trained rightly in task. The security dogs the adult escorts, facility expense, the explosive and narcotic drug detection, it will be able to use with the other blind man guidance dogs back. The narcotic drug detection dogs which currently is used specially technique intelligence anger, when considering the tendency of the narcotic drug smuggling offense field which becomes diversification that the role very it is important is a possibility of saying at day. It cultivate a fourth, escort relation specialty manpower and it improves the breed of the security dogs. The hazard which cultivate the security dogs use necessary personnel the breed of security dogs, the security dogs training center it opens the security crane relation subject of the college which stands and (university) it improves it establishes and training which is suitable in task it is to do to execute letting in the training map company. Specially, the hazard which improves the breed of security dogs in the progress mind quality which stands against the portion where the breed improvement is demanded as the portion where the internal organs research and investment are necessary sees. The security dogs compares in labor cost and the expense holds few, if it uses the our specific domestic dogs it will be able to use efficiently in the task which is various it solves the multi branch plans for wisly with the security dogs industrial development security of course contemporary history sliced raw fish sees demands compared to being immediacy and the life which is happy business the place where it does it sees it will be able to contribute a lot as.

• Journal of Drive and Control
• /
• v.17 no.3
• /
• pp.33-46
• /
• 2020

We introduce a piece of special-purpose equipment for responding to disasters that has a dual-arm manipulator consisting of six-axis multi joints, and a master-slave operating system controlled by a wearable joystick for intuitive and convenient operation. However, due to the complexity and diversity of a disaster environment, training and suitable training means are needed to improve the interaction between the driver and equipment. Therefore, in this paper, a system that can improve the operator's immersion in the training simulation is proposes, this system is implemented in a virtual environment. The implemented system consists of a cabin installed with the master-slave operation system, a motion platform, visual and sound systems, as well as a real-time simulation device. This whole system was completed by applying various techniques such as a statistical mapping method, inverse kinematics, and a real-time physical model. Then, the implemented system was evaluated from a point of view of the appropriateness of the mapping method, inverse kinematics, the feasibility for real-time simulations of the physical environment through some task mode.