• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.2
• /
• pp.115-122
• /
• 2017

This paper proposes a deep learning algorithm based sign detection and recognition system for the blind. The proposed system is composed of sign detection stage and sign recognition stage. In the sign detection stage, aggregated channel features are extracted and AdaBoost classifier is applied to detect regions of interest of the sign. In the sign recognition stage, convolutional neural network is applied to recognize the regions of interest of the sign. In this paper, the AdaBoost classifier is designed to decrease the number of undetected signs, and deep learning algorithm is used to increase recognition accuracy and which leads to removing false positives which occur in the sign detection stage. Based on our experiments, proposed method efficiently decreases the number of false positives compared with other methods.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.11
• /
• pp.1371-1376
• /
• 2019

In case of fire in a structure, it is difficult to suppress fire because it can not accurately grasp the location of fire in case of fire. In this paper, we propose a system algorithm that can guide the optimal evacuation route in case of deep learning-based (RNN) structure disaster. The present invention provides a service to transmit data detected by sensors to a server in real time by using installed sensor, to transmit and analyze information such as temperature, heat, smoke, toxic gas around the sensor, to identify the safest moving path within a set threshold, to transmit information to LED guide lights and direction indicators in a structure in real time to avoid risk factors. This is because the information of temperature, heat, smoke, and toxic gas in each area of the structure can be grasped, and it is considered that the optimal evacuation route can be guided in case of structure disaster.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.11
• /
• pp.29-38
• /
• 2022

This paper proposes a system for classifying gait types using an ensemble deep learning network for gait data measured by a smart insole equipped with multi-sensors. The gait type classification system consists of a part for normalizing the data measured by the insole, a part for extracting gait features using a deep learning network, and a part for classifying the gait type by inputting the extracted features. Two kinds of gait feature maps were extracted by independently learning networks based on CNNs and LSTMs with different characteristics. The final ensemble network classification results were obtained by combining the classification results. For the seven types of gait for adults in their 20s and 30s: walking, running, fast walking, going up and down stairs, and going up and down hills, multi-sensor data was classified into a proposed ensemble network. As a result, it was confirmed that the classification rate was higher than 90%.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2020.06a
• /
• pp.128-128
• /
• 2020

최근 농업환경의 변화와 기후변화에 대응하기 위해 농업용수 관리 정보화 및 과학화의 필요성이 증대되어 실시간으로 저수지 저수량과 농업용수 공급량을 파악하기 위해 자동 수위계측시설이 도입되었다. 농림축산식품부의 저수지 자동수위측정기 설치 및 운영지침에 따라 현재 농어촌공사 관리 저수지 1,734개소 및 수로부 1,880개소에 자동수위계가 설치되어 있으며, 저수지와 수로에서 10분 간격으로 수위자료가 생성되고 있다. 농업용 저수지 수문자료의 공인지점은 2016년 6개소에서 2019년 49개소로 증대되고 있으며, 데이터 품질 저하의 최소화 및 신뢰성 있는 수문자료 생성의 필요성이 증가함에 따라 농업용 저수지의 특성을 반영한 저수지 수위 오결측 데이터 보정 방안 및 수문 자료 품질관리 방안이 요구된다. 농업용 저수지의 수위 변화 및 강우-유출 현상은 물리적 모형을 구축하여 기상, 지형 등 영향 인자와 수위(또는 유출)와의 상관관계를 분석하는 것은 무적으로 불가능하였지만, 최근 인공신경망 (Artificial Neural Network, ANN) 등과 같이 black-box 형태의 모형을 이용하여 비선형적인 수문해석이 가능해졌다. 본 연구에서는 빅데이터와 인공신경망을 결합시킨 알고리즘인 딥러닝 (Deep Learning) 기반의 LSTM (Long Short-Term Memory) 모형을 활용하여 농업용 저수지 수위자료를 검토하여 자동계측기에서 발생하는 오류 보정을 위해 품질관리 방안을 제시하고자 한다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.1
• /
• pp.1-8
• /
• 2024

On-device AI technology, which can operate AI models at the edge devices to support real-time processing and privacy enhancement, is attracting attention. As intelligent IoT is applied to various industries, services utilizing the on-device AI technology are increasing significantly. However, general deep learning models require a lot of computational resources for inference and learning. Therefore, various lightweighting methods such as quantization and pruning have been suggested to operate deep learning models in embedded edge devices. Among the lightweighting methods, we analyze how to lightweight and apply deep learning models to edge computing devices, focusing on pruning technology in this paper. In particular, we utilize dynamic and static pruning techniques to evaluate the inference speed, accuracy, and memory usage of a lightweight AI vision model. The content analyzed in this paper can be used for intelligent video control systems or video security systems in autonomous vehicles, where real-time processing are highly required. In addition, it is expected that the content can be used more effectively in various IoT services and industries.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.2
• /
• pp.511-516
• /
• 2024

The recently enacted detailed guidelines on the performance evaluation of track facilities presented the necessary requirements regarding the evaluation procedures and implementation methods of track performance evaluation. However, the grade of rail surface damage is determined by external inspection (visual inspection), and there is no choice but to rely only on qualitative evaluation based on the subjective judgment of the inspector. Therefore, in this study, we attempted to develop a diagnostic application that can diagnose rail internal defects using rail surface damage. In the field investigation, rail surface damage was investigated and patterns were analyzed. Additionally, in the indoor test, SEM testing was used to construct image data of rail internal damage, and crack length, depth, and angle were quantified. In this study, a deep learning model (Fast R-CNN) using image data constructed from field surveys and indoor tests was applied to the application. A rail surface damage diagnosis application (App) using a deep learning model that can be used on smart devices was developed. We developed a smart diagnosis system for rail surface damage that can be used in future track diagnosis and performance evaluation work.

• Journal of the Institute of Convergence Signal Processing
• /
• v.24 no.4
• /
• pp.213-220
• /
• 2023

Falling occurs unexpectedly during daily activities, causing many difficulties in life. The purpose of this study was to establish a system for fall detection of high-risk occupations and to verify their effectiveness by collecting data and applying it to predictive models. To this end, a wearable device was configured to detect fall by calculating acceleration signals and azimuths through acceleration sensors and gyro sensors. In addition, the study participants wore the device on their abdomen and measured necessary data from falls-related movements in the process of performing predetermined activities and transmitted it to the computer through a Bluetooth device present in the device. The collected data was processed through filtering, applied to fall detection prediction models based on deep learning algorithms which are 1D CNN, LSTM and CNN-LSTM, and evaluate the results.

• Journal of Korea Society of Industrial Information Systems
• /
• v.29 no.4
• /
• pp.13-24
• /
• 2024

In this paper, we analyzed the performance of normalization according to various neural network models and input-output structures. For the analysis, a simulation-based dataset for noise environments with homogeneous and up to three interfering signals was used. As a result, the end-to-end structure that directly outputs noise variance showed superior performance when using a 1-D convolutional neural network and BiLSTM model, and was analyzed to be particularly robust against interference signals. This is because the 1-D convolutional neural network and bidirectional long short-term memory models have stronger inductive bias than the multilayer perceptron and transformer models. The analysis of this paper are expected to be used as a useful reference for future research on deep learning-based normalization.

• Annual Conference of KIPS
• /
• 2022.11a
• /
• pp.758-760
• /
• 2022

본 연구는 머신러닝의 한 종류인 YOLOv5를 이용하여 기존 육안 선별작업을 자동화 하는 기계를 설계하는 것이다. 본 연구에서는 영상촬영과 선별작업을 진행하는 컨베이어 기구와 선별 프로그램을 제작하고, 모든 표면을 검사해 사과의 품질을 3단계로 구별하는 작업을 진행하였다. 결과적으로 투입된 사과의 품질을 성공적으로 분류 하였다.

• Annual Conference on Human and Language Technology
• /
• 2017.10a
• /
• pp.116-121
• /
• 2017

cQA(Community-based Question Answering) 시스템은 온라인 커뮤니티를 통해 사용자들이 질문을 남기고 답변을 작성할 수 있도록 만들어진 시스템이다. 신규 질문이 인입되면, 기존에 축적된 cQA 저장소에서 해당 질문과 가장 유사한 질문을 검색하고, 그 질문에 대한 답변을 신규 질문에 대한 답변으로 대체할 수 있다. 하지만, 키워드 매칭을 사용하는 전통적인 검색 방식으로는 문장에 내재된 의미들을 이용할 수 없다는 한계가 있다. 이를 극복하기 위해서는 의미적으로 동일한 문장들로 학습이 되어야 하지만, 이러한 데이터를 대량으로 확보하기에는 어려움이 있다. 본 논문에서는 질문이 제목과 내용으로 분리되어 있는 대량의 cQA 셋에서, 질문 제목과 내용을 의미 벡터 공간으로 사상하고 두 벡터의 상대적 거리가 가깝게 되도록 학습함으로써 의사(pseudo) 유사 의미의 성질을 내재화 하였다. 또한, 질문 제목과 내용의 의미 벡터 표현(representation)을 위하여, semi-training word embedding과 CNN(Convolutional Neural Network)을 이용한 딥러닝 기법을 제안하였다. 유사 질문 검색 실험 결과, 제안 모델을 이용한 검색이 키워드 매칭 기반 검색보다 좋은 성능을 보였다.