• Journal of the Korean Geotechnical Society
• /
• v.39 no.10
• /
• pp.41-48
• /
• 2023

The presence of abnormalities in the subgrade of roads poses safety risks to users and results in significant maintenance costs. In this study, we aimed to experimentally evaluate the temperature distributions in abnormal areas of subgrade materials using infrared cameras and analyze the data with machine learning techniques. The experimental site was configured as a cubic shape measuring 50 cm in width, length, and depth, with abnormal areas designated for water and air. Concrete blocks covered the upper part of the site to simulate the pavement layer. Temperature distribution was monitored over 23 h, from 4 PM to 3 PM the following day, resulting in image data and numerical temperature values extracted from the middle of the abnormal area. The temperature difference between the maximum and minimum values measured 34.8℃ for water, 34.2℃ for air, and 28.6℃ for the original subgrade. To classify conditions in the measured images, we employed the image analysis method of a convolutional neural network (CNN), utilizing ResNet-101 and SqueezeNet networks. The classification accuracies of ResNet-101 for water, air, and the original subgrade were 70%, 50%, and 80%, respectively. SqueezeNet achieved classification accuracies of 60% for water, 30% for air, and 70% for the original subgrade. This study highlights the effectiveness of CNN algorithms in analyzing subgrade properties and predicting subsurface conditions.

• Journal of Bio-Environment Control
• /
• v.33 no.1
• /
• pp.1-11
• /
• 2024

In this study, five artificial intelligence (AI) models: Inception v3, SqueezeNet (local), VGG-16, Painters, and DeepLoc were used to classify tea leaf diseases. Eight image categories were used: healthy, algal leaf spot, anthracnose, bird's eye spot, brown blight, gray blight, red leaf spot, and white spot. Software used in this study was Orange 3 which functions as a Python library for visual programming, that operates through an interface that generates workflows to visually manipulate and analyze the data. The precision of each AI model was recorded to select the ideal AI model. All models were trained using the Adam solver, rectified linear unit activation function, 100 neurons in the hidden layers, 200 maximum number of iterations in the neural network, and 0.0001 regularizations. To extend the functionality of Orange 3, new add-ons can be installed and, this study image analytics add-on was newly added which is required for image analysis. For the training model, the import image, image embedding, neural network, test and score, and confusion matrix widgets were used, whereas the import images, image embedding, predictions, and image viewer widgets were used for the prediction. Precisions of the neural networks of the five AI models (Inception v3, SqueezeNet (local), VGG-16, Painters, and DeepLoc) were 0.807, 0.901, 0.780, 0.800, and 0.771, respectively. Finally, the SqueezeNet (local) model was selected as the optimal AI model for the detection of tea diseases using tea leaf images owing to its high precision and good performance throughout the confusion matrix.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.226-227
• /
• 2020

근래의 초해상화 (super-resolution, SR) 연구는 네트워크를 깊고, 넓게 만들어 성능을 높이는데 주를 이뤘다. 그러나 동시에 높은 연산량과 메모리 소비량이 증가하는 문제가 발생하기 때문에 이를 실제로 하드웨어로 구현하기에는 어려운 문제가 존재한다. 그렇기에 우리는 네트워크 최적화를 통해 성능 감소를 최소화하면서 파라미터 수를 줄이는 네트워크 SqueezeSR을 설계하였다. 또한 지식 증류(Knowledge Distillation, KD)를 이용해 추가적인 파라미터 수 증가 없이 성능을 높일 수 있는 학습 방법을 제안한다. 또한 KD 시 teacher network의 성능이 보다 student network에 잘 전달되도록 feature map 간의 비교를 통해 학습 효율을 높일 수 있었다. 결과적으로 우리는 KD 기법을 통해 추가적인 파라미터 수 증가 없이 성능을 높여 다른 SR네트워크 대비 더 빠르고 성능 감소를 최소화한 네트워크를 제안한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.10a
• /
• pp.631-633
• /
• 2018

최근 영상처리 분야에서 딥러닝(Deep learning)을 이용한 기술이 좋은 성능을 보이면서 이에 대한 관심과 연구가 증가하고 있다. 본 연구에서는 최근 딥러닝 네트워크 중 적은 파라미터 수로 AlexNet수준의 성능을 보인 SquezeNet을 영상 분할(Image segmentation)의 특징 추출(feature extraction)영역으로 사용하고, CUDA C기반으로 코드를 작성하여 정확도를 유지하면서 계산 속도 면에서도 좋은 성능을 얻을 수 있었다.

• Journal of Korea Foundry Society
• /
• v.18 no.4
• /
• pp.373-382
• /
• 1998

Particulate reinforced aluminum alloys produced by indirect squeeze casting are difficult to shape by cutting or milling. Therefore near net shape forming of complex shapes is of high economic and technical interest. The complex shape products of $SiC_p/6061$ Al composites are fabricated by the melt-stirring and indirect squeeze casting process. The mold temperatures are $200^{\circ}C$ and $300^{\circ}C$ and applied pressures are 70, 100, and 130 MPa. The volume fractions of the reinforcements are in the range of 5 vol% to 15 vol%. The reinforcement dispersion state are observed using on optical microscope. By employing observed results systematically a correlation is demonstrated among the microstructure, particles behavior, mechanical properties and processing parameters for an optimum melt-stirring(compocasting) and indirect squeeze casting process of MMCs. A procedure to establish the optimum squeeze casting of Al-MMCs is proposed.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• fall
• /
• pp.154-156
• /
• 2021

Depth-wise separable convolution 은 컴퓨터 자원이 제한된 환경에서 기존의 standard convolution을 대체하는데 강력하고, 효과적인 대안으로 잘 알려져 있다.[1] MobileNetV2 에서는 Inverted residual block을 소개한다. 이는 depth-wise separable convolution으로 인해 생기는 손실, 즉 channel 간의 데이터를 조합해 새로운 feature를 만들어낼 기회를 잃어버릴 때, 이를 depth-wise separable convolution 양단에 point-wise convolution(1×1 convolution)을 사용함으로써 극복해낸 block이다.[1] 하지만 1×1 convolution은 채널 수에 의존적(dependent)인 특징을 갖고 있고, 따라서 결국 네트워크가 깊어지면 깊어질수록 효율적이고(efficient) 가벼운(light weight) 네트워크를 만드는데 병목 현상(bottleneck)을 일으키고 만다. 이 논문에서는 channel-wise squeeze and excitation block(CSE)을 통해 1×1 convolution을 부분적으로 대체하는 방법을 통해 이 병목 현상을 해결한다.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.17 no.7 s.94
• /
• pp.1751-1760
• /
• 1993

The squeeze casting process is considered as an attractive way to form the primary product of near net shape metal matrix composites for wide use in automobile industry. To understand for infiltration limit in squeeze casting processes, the SAFFIL short fiber preform of volume fraction $10%{\sim}23%$ were fabricated by vaccum pumping and speed control press, and the optimal condition for fiber preform fabrication had been experimentally obtained. The composite materials were fabricated by forced infiltration of molten metals such as Al6061, Al7075, pure Al, AC8A, and Al2024. The infiltration distance and deformation of fiber preform are observed, and tensile strength were measured from at the room temperature.

• Korean Journal of Radiology
• /
• v.20 no.8
• /
• pp.1275-1284
• /
• 2019

Objective: To develop algorithms using convolutional neural networks (CNNs) for automatic segmentation of acute ischemic lesions on diffusion-weighted imaging (DWI) and compare them with conventional algorithms, including a thresholding-based segmentation. Materials and Methods: Between September 2005 and August 2015, 429 patients presenting with acute cerebral ischemia (training:validation:test set = 246:89:94) were retrospectively enrolled in this study, which was performed under Institutional Review Board approval. Ground truth segmentations for acute ischemic lesions on DWI were manually drawn under the consensus of two expert radiologists. CNN algorithms were developed using two-dimensional U-Net with squeeze-and-excitation blocks (U-Net) and a DenseNet with squeeze-and-excitation blocks (DenseNet) with squeeze-and-excitation operations for automatic segmentation of acute ischemic lesions on DWI. The CNN algorithms were compared with conventional algorithms based on DWI and the apparent diffusion coefficient (ADC) signal intensity. The performances of the algorithms were assessed using the Dice index with 5-fold cross-validation. The Dice indices were analyzed according to infarct volumes (< 10 mL, ≥ 10 mL), number of infarcts (≤ 5, 6-10, ≥ 11), and b-value of 1000 (b1000) signal intensities (< 50, 50-100, > 100), time intervals to DWI, and DWI protocols. Results: The CNN algorithms were significantly superior to conventional algorithms (p < 0.001). Dice indices for the CNN algorithms were 0.85 for U-Net and DenseNet and 0.86 for an ensemble of U-Net and DenseNet, while the indices were 0.58 for ADC-b1000 and b1000-ADC and 0.52 for the commercial ADC algorithm. The Dice indices for small and large lesions, respectively, were 0.81 and 0.88 with U-Net, 0.80 and 0.88 with DenseNet, and 0.82 and 0.89 with the ensemble of U-Net and DenseNet. The CNN algorithms showed significant differences in Dice indices according to infarct volumes (p < 0.001). Conclusion: The CNN algorithm for automatic segmentation of acute ischemic lesions on DWI achieved Dice indices greater than or equal to 0.85 and showed superior performance to conventional algorithms.

• ETRI Journal
• /
• v.45 no.2
• /
• pp.318-328
• /
• 2023

Recently, embedded systems, such as mobile platforms, have multiple processing units that can operate in parallel, such as centralized processing units (CPUs) and neural processing units (NPUs). We can use deep-learning compilers to generate machine code optimized for these embedded systems from a deep neural network (DNN). However, the deep-learning compilers proposed so far generate codes that sequentially execute DNN operators on a single processing unit or parallel codes for graphic processing units (GPUs). In this study, we propose PartitionTuner, an operator scheduler for deep-learning compilers that supports multiple heterogeneous PUs including CPUs and NPUs. PartitionTuner can generate an operator-scheduling plan that uses all available PUs simultaneously to minimize overall DNN inference time. Operator scheduling is based on the analysis of DNN architecture and the performance profiles of individual and group operators measured on heterogeneous processing units. By the experiments for seven DNNs, PartitionTuner generates scheduling plans that perform 5.03% better than a static type-based operator-scheduling technique for SqueezeNet. In addition, PartitionTuner outperforms recent profiling-based operator-scheduling techniques for ResNet50, ResNet18, and SqueezeNet by 7.18%, 5.36%, and 2.73%, respectively.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.5
• /
• pp.1431-1445
• /
• 2022

We construct a deep neural network model named ECGResNet. This model can diagnosis diseases based on 12-lead ECG data of eight common cardiovascular diseases with a high accuracy. We chose the 16 Blocks of ResNet50 as the main body of the model and added the Squeeze-and-Excitation module to learn the data information between channels adaptively. We modified the first convolutional layer of ResNet50 which has a convolutional kernel of 7 to a superposition of convolutional kernels of 8 and 16 as our feature extraction method. This way allows the model to focus on the overall trend of the ECG signal while also noticing subtle changes. The model further improves the accuracy of cardiovascular and cerebrovascular disease classification by using a fully connected layer that integrates factors such as gender and age. The ECGResNet model adds Dropout layers to both the residual block and SE module of ResNet50, further avoiding the phenomenon of model overfitting. The model was eventually trained using a five-fold cross-validation and Flooding training method, with an accuracy of 95% on the test set and an F1-score of 0.841.We design a new deep neural network, innovate a multi-scale feature extraction method, and apply the SE module to extract features of ECG data.