• 한국재료학회지
• /
• 제32권8호
• /
• pp.345-353
• /
• 2022

In this study, using deep learning, super-resolution images of transmission electron microscope (TEM) images were generated for nanomaterial analysis. 1169 paired images with 256 × 256 pixels (high resolution: HR) from TEM measurements and 32 × 32 pixels (low resolution: LR) produced using the python module openCV were trained with deep learning models. The TEM images were related to DyVO₄ nanomaterials synthesized by hydrothermal methods. Mean-absolute-error (MAE), peak-signal-to-noise-ratio (PSNR), and structural similarity (SSIM) were used as metrics to evaluate the performance of the models. First, a super-resolution image (SR) was obtained using the traditional interpolation method used in computer vision. In the SR image at low magnification, the shape of the nanomaterial improved. However, the SR images at medium and high magnification failed to show the characteristics of the lattice of the nanomaterials. Second, to obtain a SR image, the deep learning model includes a residual network which reduces the loss of spatial information in the convolutional process of obtaining a feature map. In the process of optimizing the deep learning model, it was confirmed that the performance of the model improved as the number of data increased. In addition, by optimizing the deep learning model using the loss function, including MAE and SSIM at the same time, improved results of the nanomaterial lattice in SR images were achieved at medium and high magnifications. The final proposed deep learning model used four residual blocks to obtain the characteristic map of the low-resolution image, and the super-resolution image was completed using Upsampling2D and the residual block three times.

• Journal of Positioning, Navigation, and Timing
• /
• 제11권3호
• /
• pp.217-227
• /
• 2022

In order to provide a location-based services regardless of indoor or outdoor space, it is important to provide position information of the terminal regardless of location. Among the wireless/mobile communication resources used for this purpose, Long Term Evolution (LTE) signal is a representative infrastructure that can overcome spatial limitations, but the positioning method based on the location of the base station has a disadvantage in that the accuracy is low. Therefore, a fingerprinting technique, which is a pattern recognition technology, has been widely used. The simplest yet widely applied algorithm among Fingerprint positioning technologies is k-Nearest Neighbors (kNN). However, in the kNN algorithm, it is difficult to find the optimal K value with the lowest positioning error for each location to be estimated, so it is generally fixed to an appropriate K value and used. Since the optimal K value cannot be applied to each estimated location, therefore, there is a problem in that the accuracy of the overall estimated location information is lowered. Considering this problem, this paper proposes a technique for adaptively varying the K value by using a Convolutional Neural Network (CNN) model among Artificial Neural Network (ANN) techniques. First, by using the signal information of the measured values obtained in the service area, an image is created according to the Physical Cell Identity (PCI) and Band combination, and an answer label for supervised learning is created. Then, the structure of the CNN is modeled to classify K values through the image information of the measurements. The performance of the proposed technique is verified based on actual data measured in the testbed. As a result, it can be seen that the proposed technique improves the positioning performance compared to using a fixed K value.

• 방송공학회논문지
• /
• 제28권2호
• /
• pp.178-184
• /
• 2023

카메라로 촬영한 야외 일반 영상에서 텍스트 이미지를 찾아내고 그 내용을 인식하는 기술은 로봇 비전, 시각 보조 등의 기반으로 활용될 수 있는 매우 중요한 기술이다. 하지만 텍스트 이미지가 저해상도인 경우에는 텍스트 이미지에 포함된 노이즈나 블러 등의 열화가 더 두드러지기 때문에 텍스트 내용 인식 성능의 하락이 발생하게 된다. 본 논문에서는 일반 영상에서의 저해상도 한글 텍스트에 대한 이미지 초해상화를 통해서 텍스트 인식 정확도를 개선하였다. 트랜스포머에 기반한 모델로 한글 텍스트 이미지 초해상화를 수행 하였으며, 직접 구축한 고해상도-저해상도 한글 텍스트 이미지 데이터셋에 대하여 제안한 초해상화 방법을 적용했을 때 텍스트 인식 성능이 개선되는 것을 확인하였다.

• Imaging Science in Dentistry
• /
• 제52권4호
• /
• pp.351-357
• /
• 2022

Purpose: Convolutional neural networks (CNNs) have rapidly emerged as one of the most promising artificial intelligence methods in the field of medical and dental research. CNNs can provide an effective diagnostic methodology allowing for the detection of early-staged diseases. Therefore, this study aimed to evaluate the performance of a deep CNN algorithm for apical lesion segmentation from panoramic radiographs. Materials and Methods: A total of 1000 panoramic images showing apical lesions were separated into training (n=800, 80%), validation (n=100, 10%), and test (n=100, 10%) datasets. The performance of identifying apical lesions was evaluated by calculating the precision, recall, and F1-score. Results: In the test group of 180 apical lesions, 147 lesions were segmented from panoramic radiographs with an intersection over union (IoU) threshold of 0.3. The F1-score values, as a measure of performance, were 0.828, 0.815, and 0.742, respectively, with IoU thresholds of 0.3, 0.4, and 0.5. Conclusion: This study showed the potential utility of a deep learning-guided approach for the segmentation of apical lesions. The deep CNN algorithm using U-Net demonstrated considerably high performance in detecting apical lesions.

• 대한의용생체공학회:의공학회지
• /
• 제44권2호
• /
• pp.147-156
• /
• 2023

Efforts to employ smart home sensors to monitor the indoor activities of elderly single residents have been made to assess the feasibility of a safe and healthy lifestyle. However, the bathroom remains an area of blind spot. In this study, we have developed and evaluated a new edge computer device that can automatically detect water usage activities in the bathroom and record the activity log on a cloud server. Three kinds of sound as flushing, showering, and washing using wash basin generated during water usage were recorded and cut into 1-second scenes. These sound clips were then converted into a 2-dimensional image using MEL-spectrogram. Sound data augmentation techniques were adopted to obtain better learning effect from smaller number of data sets. These techniques, some of which are applied in time domain and others in frequency domain, increased the number of training data set by 30 times. A deep learning model, called CRNN, combining Convolutional Neural Network and Recurrent Neural Network was employed. The edge device was implemented using Raspberry Pi 4 and was equipped with a condenser microphone and amplifier to run the pre-trained model in real-time. The detected activities were recorded as text-based activity logs on a Firebase server. Performance was evaluated in two bathrooms for the three water usage activities, resulting in an accuracy of 96.1% and 88.2%, and F1 Score of 96.1% and 87.8%, respectively. Most of the classification errors were observed in the water sound from washing. In conclusion, this system demonstrates the potential for use in recording the activities as a lifelog of elderly single residents to a cloud server over the long-term.

• Smart Structures and Systems
• /
• 제31권4호
• /
• pp.351-363
• /
• 2023

Visual structural inspections are an inseparable part of post-earthquake damage assessments. With unmanned aerial vehicles (UAVs) establishing a new frontier in visual inspections, there are major computational challenges in processing the collected massive amounts of high-resolution visual data. We propose twin deep learning models that can provide accurate high-resolution structural components and damage segmentation masks efficiently. The traditional approach to cope with high memory computational demands is to either uniformly downsample the raw images at the price of losing fine local details or cropping smaller parts of the images leading to a loss of global contextual information. Therefore, our twin models comprising Trainable Resizing for high-resolution Segmentation Network (TRS-Net) and DmgFormer approaches the global and local semantics from different perspectives. TRS-Net is a compound, high-resolution segmentation architecture equipped with learnable downsampler and upsampler modules to minimize information loss for optimal performance and efficiency. DmgFormer utilizes a transformer backbone and a convolutional decoder head with skip connections on a grid of crops aiming for high precision learning without downsizing. An augmented inference technique is used to boost performance further and reduce the possible loss of context due to grid cropping. Comprehensive experiments have been performed on the 3D physics-based graphics models (PBGMs) synthetic environments in the QuakeCity dataset. The proposed framework is evaluated using several metrics on three segmentation tasks: component type, component damage state, and global damage (crack, rebar, spalling). The models were developed as part of the 2^nd International Competition for Structural Health Monitoring.

• 대한임베디드공학회논문지
• /
• 제18권3호
• /
• pp.89-100
• /
• 2023

Recently, on-device artificial intelligence (AI) solutions using mobile devices and embedded edge devices have emerged in various fields, such as computer vision, to address network traffic burdens, low-energy operations, and security problems. Although vision transformer deep learning models have outperformed conventional convolutional neural network (CNN) models in computer vision, they require more computations and parameters than CNN models. Thus, they are not directly applicable to embedded edge devices with limited hardware resources. Many researchers have proposed various model compression methods or lightweight architectures for vision transformers; however, there are only a few studies evaluating the effects of model compression techniques of vision transformers on performance. Regarding this problem, this paper presents a performance evaluation of vision transformers on embedded platforms. We investigated the behaviors of three vision transformers: DeiT, LeViT, and MobileViT. Each model performance was evaluated by accuracy and inference time on edge devices using the ImageNet dataset. We assessed the effects of the quantization method applied to the models on latency enhancement and accuracy degradation by profiling the proportion of response time occupied by major operations. In addition, we evaluated the performance of each model on GPU and EdgeTPU-based edge devices. In our experimental results, LeViT showed the best performance in CPU-based edge devices, and DeiT-small showed the highest performance improvement in GPU-based edge devices. In addition, only MobileViT models showed performance improvement on EdgeTPU. Summarizing the analysis results through profiling, the degree of performance improvement of each vision transformer model was highly dependent on the proportion of parts that could be optimized in the target edge device. In summary, to apply vision transformers to on-device AI solutions, either proper operation composition and optimizations specific to target edge devices must be considered.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2023년도 학술발표회
• /
• pp.65-65
• /
• 2023

최근 기후변화로 인해 홍수, 가뭄 등 수재해가 세계 곳곳에서 빈번하게 발생하고 있다. 이로 인해 정확한 강우-유출 해석의 중요도는 높아지고 있으며 강우-유출 해석에 따라 수자원 관리 및 계획수립의 정도가 달라질 수 있다. 본 연구 대상 지역인 메콩강 유역은 중국과 동남아시아 5개국(라오스, 태국, 미얀마, 베트남, 캄보디아)을 관통하는 국가공유하천으로 기초자료의 획득이 어렵고 국가별로 구축된 자료가 질적, 양적 품질이 상이하여 수문해석에서의 기초자료로 사용하기에 불확실성이 있다. 최근 기술의 발달로 글로벌 격자형 강수자료 획득이 용이함에 있어 미계측 대유역에서의 다양한 연구들이 수행되고 있지만, 지점강수자료와 시·공간적 오차로 인한 불확실성을 내포하고 있다. 이에 본 연구에서는 글로벌 격자형 강수자료의 적용성을 평가하기 위하여 지점 격자형 강수자료(APHRODITE)와 4개의 위성강수자료(CHIRPS, CMORPH, PERSIANN-CDR, TRMM)를 수집하고 합성곱 신경망 모형인 ConvAE 기법을 이용하여 위성강수자료의 시·공간 편의 보정을 수행하였다. 또한, 하천 수위에 대한 장기간 정보 수집이 가능한 메콩강 본류 4개 관측소(Luang Prabang, Pakse, Stung Treng, Kratie)를 선정하였으며 SWAT 모형을 이용하여 매개변수 보정(2004~2013)과 격자형 강수자료의 보정 전·후의 유출모의(2014~2015) 결과를 비교·분석하였다. 격자형 강우를 이용한 보정 및 유출 분석 결과 4개의 위성강수자료 모두 성능이 향상되었으며 그 중 보정된 TRMM이 가장 우수한 성능을 보여 해당 유역에서의 APHRODITE를 대체할 수 있다고 판단하였다. 따라서 본 연구에서 제시하는 ConvAE를 이용한 보정기법과 이를 이용한 강우-유출 해석은 향후 다양한 격자형 강수자료를 활용한 미계측 대유역에서의 수문해석에서 활용이 가능할 것으로 판단된다.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2023년도 학술발표회
• /
• pp.358-358
• /
• 2023

본 연구에서는 그동안 수자원분야 강우유출 해석분야에 활용되었던 대표적인 머신러닝&딥러닝(ML&DL) 모델을 활용하여 모델의 하이퍼파라미터 튜닝뿐만 아니라 모델의 특성을 고려한 기상 및 수문데이터의 조합과 전처리(lag-time, 이동평균 등)를 통하여 데이터 특성과 ML&DL모델의 조합시나리오에 따른 일 유입량 예측성능을 비교 검토하는 연구를 수행하였다. 이를 위해 소양강댐 유역을 대상으로 1974년에서 2021년까지 축적된 기상 및 수문데이터를 활용하여 1) 강우, 2) 유입량, 3) 기상자료를 주요 영향변수(독립변수)로 고려하고, 이에 a) 지체시간(lag-time), b) 이동평균, c) 유입량의 성분분리조건을 적용하여 총 36가지 시나리오 조합을 ML&DL의 입력자료로 활용하였다. ML&DL 모델은 1) Linear Regression(LR), 2) Lasso, 3) Ridge, 4) SVR(Support Vector Regression), 5) Random Forest(RF), 6) LGBM(Light Gradient Boosting Model), 7) XGBoost의 7가지 ML방법과 8) LSTM(Long Short-Term Memory models), 9) TCN(Temporal Convolutional Network), 10) LSTM-TCN의 3가지 DL 방법, 총 10가지 ML&DL모델을 비교 검토하여 일유입량 예측을 위한 가장 적합한 데이터 조합 특성과 ML&DL모델을 성능평가와 함께 제시하였다. 학습된 모형의 유입량 예측 결과를 비교·분석한 결과, 소양강댐 유역에서는 딥러닝 중에서는 TCN모형이 가장 우수한 성능을 보였고(TCN>TCN-LSTM>LSTM), 트리기반 머신러닝중에서는 Random Forest와 LGBM이 우수한 성능을 보였으며(RF, LGBM>XGB), SVR도 LGBM수준의 우수한 성능을 나타내었다. LR, Lasso, Ridge 세가지 Regression모형은 상대적으로 낮은 성능을 보였다. 또한 소양강댐 댐유입량 예측에 대하여 강우, 유입량, 기상계열을 36가지로 조합한 결과, 입력자료에 lag-time이 적용된 강우계열의 조합 분석에서 세가지 Regression모델을 제외한 모든 모형에서 NSE(Nash-Sutcliffe Efficiency) 0.8이상(최대 0.867)의 성능을 보였으며, lag-time이 적용된 강우와 유입량계열을 조합했을 경우 NSE 0.85이상(최대 0.901)의 더 우수한 성능을 보였다.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2023년도 학술발표회
• /
• pp.173-173
• /
• 2023

The Mekong River Basin (MRB) is a crucial watershed in Asia, impacting over 60 million people across six developing nations. Accurate satellite-based precipitation products (SPPs) are essential for effective hydrological and watershed management in this region. However, the performance of SPPs has been varied and limited. The APHRODITE product, a unique gauge-based dataset for MRB, is widely used but is only available until 2015. In this study, we present a novel framework for correcting SPPs in the MRB by employing a deep learning approach that combines convolutional neural networks and encoder-decoder architecture to address pixel-by-pixel bias and enhance accuracy. The DLF was applied to four widely used SPPs (TRMM, CMORPH, CHIRPS, and PERSIANN-CDR) in MRB. For the original SPPs, the TRMM product outperformed the other SPPs. Results revealed that the DLF effectively bridged the spatial-temporal gap between the SPPs and the gauge-based dataset (APHRODITE). Among the four corrected products, ADJ-TRMM demonstrated the best performance, followed by ADJ-CDR, ADJ-CHIRPS, and ADJ-CMORPH. The DLF offered a robust and adaptable solution for bias correction in the MRB and beyond, capable of detecting intricate patterns and learning from data to make appropriate adjustments. With the discontinuation of the APHRODITE product, DLF represents a promising solution for generating a more current and reliable dataset for MRB research. This research showcased the potential of deep learning-based methods for improving the accuracy of SPPs, particularly in regions like the MRB, where gauge-based datasets are limited or discontinued.