• The Journal of Korea Robotics Society
• /
• v.17 no.3
• /
• pp.264-272
• /
• 2022

Human-robot cooperative tasks are increasingly required in our daily life with the development of robotics and artificial intelligence technology. Interactive reinforcement learning strategies suggest that robots learn task by receiving feedback from an experienced human trainer during a training process. However, most of the previous studies on Interactive reinforcement learning have required an extra feedback input device such as a mouse or keyboard in addition to robot itself, and the scenario where a robot can interactively learn a task with human have been also limited to virtual environment. To solve these limitations, this paper studies training strategies of robot that learn table balancing tasks interactively using deep reinforcement learning with human's facial expression feedback. In the proposed system, the robot learns a cooperative table balancing task using Deep Q-Network (DQN), which is a deep reinforcement learning technique, with human facial emotion expression feedback. As a result of the experiment, the proposed system achieved a high optimal policy convergence rate of up to 83.3% in training and successful assumption rate of up to 91.6% in testing, showing improved performance compared to the model without human facial expression feedback.

• Investigative Magnetic Resonance Imaging
• /
• v.25 no.4
• /
• pp.300-312
• /
• 2021

Compressed sensing (CS) has been investigated in magnetic resonance (MR) parametric mapping to reduce scan time. However, the relatively long reconstruction time restricts its widespread applications in the clinic. Recently, deep learning-based methods have shown great potential in accelerating reconstruction time and improving imaging quality in fast MR imaging, although their adaptation to parametric mapping is still in an early stage. In this paper, we proposed a novel deep learning-based framework DEMO for fast and robust MR parametric mapping. Different from current deep learning-based methods, DEMO trains the network in an unsupervised way, which is more practical given that it is difficult to acquire large fully sampled training data of parametric-weighted images. Specifically, a CS-based loss function is used in DEMO to avoid the necessity of using fully sampled k-space data as the label, thus making it an unsupervised learning approach. DEMO reconstructs parametric weighted images and generates a parametric map simultaneously by unrolling an interaction approach in conventional fast MR parametric mapping, which enables multi-tasking learning. Experimental results showed promising performance of the proposed DEMO framework in quantitative MR T1ρ mapping.

• Earthquakes and Structures
• /
• v.25 no.6
• /
• pp.469-479
• /
• 2023

A probabilistic seismic damage analysis is an essential procedure to identify seismically vulnerable structures, prioritize the seismic retrofit, and ultimately minimize the overall seismic risk. To assess the seismic risk of multiple structures within a region, a large number of nonlinear time-history structural analyses must be conducted and studied. As a result, each assessment requires high computing resources. To overcome this limitation, we explore a deep learning-based metamodel to enable the prediction of the mean and the standard deviation of the seismic damage distribution of track-on steel-plate girder railway bridges in Korea considering the geometric variation. For machine learning training, nonlinear dynamic time-history analyses are performed to generate 800 high-fidelity datasets on the seismic response. Through intensive trial and error, the study is concentrated on developing an optimal machine learning architecture with the pre-identified variables of the physical configuration of the bridge. Additionally, the prediction performance of the proposed method is compared with a previous, well-defined, response surface model. Finally, the statistical testing results indicate that the overall performance of the deep-learning model is improved compared to the response surface model, as its errors are reduced by as much as 61%. In conclusion, the model proposed in this study can be effectively deployed for the seismic fragility and risk assessment of a region with a large number of structures.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2023.11a
• /
• pp.230-231
• /
• 2023

This study proposes a deep reinforcement learning-based algorithm to automatically generate a ship's passage plan. First, Busan Port and Gwangyang Port were selected as target areas, and a container ship with a draft of 16m was designated as the target vessel. The experimental results showed that the ship's passage plan generated using deep reinforcement learning was more efficient than the Q-learning-based algorithm used in previous research. This algorithm presents a method to generate a ship's passage plan automatically and can contribute to improving maritime safety and efficiency.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.6
• /
• pp.1187-1194
• /
• 2020

We presented a method of deep quiz cropping for automatic construction of quiz pool in online quiz systems. The method detects question boxes and sunda boxes in images captured from test papers by a deep learning-based object detector, and makes pairs of question box and sunda box by the box coupling. We applied the deep quiz cropping to images captured from test papers and achieved successful results.

• KIISE Transactions on Computing Practices
• /
• v.23 no.7
• /
• pp.417-423
• /
• 2017

Recently, many embedded devices that have the computing capability required for deep learning have become available; hence, many new applications using these devices are emerging. However, these embedded devices have an architecture different from that of PCs and high-performance servers. In this paper, we propose a method that improves the performance of deep-learning framework by considering the architecture of an embedded device that shares memory between the CPU and the GPU. The proposed method is implemented in Caffe, an open-source deep-learning framework, and is evaluated on an NVIDIA Jetson TK1 embedded device. In the experiment, we investigate the image recognition performance of several state-of-the-art deep-learning networks, including AlexNet, VGGNet, and GoogLeNet. Our results show that the proposed method can achieve significant performance gain. For instance, in AlexNet, we could reduce image recognition latency by about 33% and energy consumption by about 50%.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.6
• /
• pp.177-186
• /
• 2020

Recent advances in deep learning technology have improved image recognition performance in the field of computer vision, and serverless computing is emerging as the next generation cloud computing technology for event-based cloud application development and services. Attempts to use deep learning and serverless computing technology to increase the number of real-world image recognition services are increasing. Therefore, this paper describes how to develop an efficient deep learning based image recognition service system using serverless computing technology. The proposed system suggests a method that can serve large neural network model to users at low cost by using AWS Lambda Server based on serverless computing. We also show that we can effectively build a serverless computing system that uses a large neural network model by addressing the shortcomings of AWS Lambda Server, cold start time and capacity limitation. Through experiments, we confirmed that the proposed system, using AWS Lambda Serverless Computing technology, is efficient for servicing large neural network models by solving processing time and capacity limitations as well as cost reduction.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.17 no.4
• /
• pp.150-158
• /
• 2018

In this paper, we introduce an electric vehicle charging system using power line communication and propose a method to correct the error by applying a deep learning algorithm when an error occurs in the control signal of an electric vehicle charging system using power line communication. The error detection and correction of the control signal can be solved through the conventional error correcting code schemes, but the error is detected and corrected more efficiently by using the deep learning based error correcting code scheme. Therefore, we introduce deep learning based error correction code scheme and apply this scheme to electric vehicle charging system using power line communication. we proceed simulation and confirm performance with bit error rate. we judge whether the deep learning based error correction code scheme is more effective than the conventional schemes.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.17 no.4
• /
• pp.14-25
• /
• 2018

In recent years, it has become technically easier to explain factors related with traffic accidents in the Big Data era. Therefore, it is necessary to apply the latest analysis techniques to analyze the traffic accident data and to seek for new findings. The purpose of this study is to compare the predictive performance of the negative binomial regression model and the deep learning method developed in this study to predict the frequency of traffic accidents in expressways. As a result, the MOEs of the deep learning model are somewhat superior to those of the negative binomial regression model in terms of prediction performance. However, using a deep learning model could increase the predictive reliability. However, it is easy to add other independent variables when using deep learning, and it can be expected to increase the predictive reliability even if the model structure is changed.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2015.05a
• /
• pp.421-422
• /
• 2015

Deep learning, a deepen neural network technology that demonstrates the enhanced performance of neural network analysis, has been getting the spotlight in recent years. The present study proposed a process to test the error rates of certain variables and predict big data by using R, a analysis visualization tool based on deep learning, applying the RBM(Restricted Boltzmann Machine) algorithm to deep learning. The weighted value of each dependent variable was also applied after the classification of dependent variables. The investigator tested input data with the RBM algorithm and designed a process to detect error rates with the application of R.