• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.05a
• /
• pp.373-376
• /
• 2021

Deep neural networks have achieved almost human-level results in various tasks and have become popular in the broad artificial intelligence domains. Uncertainty estimation is an on-demand task caused by the black-box point estimation behavior of deep learning. The deep ensemble provides increased accuracy and estimated uncertainty; however, linearly increasing the size makes the deep ensemble unfeasible for memory-intensive tasks. To address this problem, we used model pruning and quantization with a deep ensemble and analyzed the effect in the context of uncertainty metrics. We empirically showed that the ensemble members' disagreement increases with pruning, making models sparser by zeroing irrelevant parameters. Increased disagreement implies increased uncertainty, which helps in making more robust predictions. Accordingly, an energy-efficient compressed deep ensemble is appropriate for memory-intensive and uncertainty-aware tasks.

• Korean Journal of Radiology
• /
• v.21 no.1
• /
• pp.33-41
• /
• 2020

Artificial intelligence has been applied to many industries, including medicine. Among the various techniques in artificial intelligence, deep learning has attained the highest popularity in medical imaging in recent years. Many articles on deep learning have been published in radiologic journals. However, radiologists may have difficulty in understanding and interpreting these studies because the study methods of deep learning differ from those of traditional radiology. This review article aims to explain the concepts and terms that are frequently used in deep learning radiology articles, facilitating general radiologists' understanding.

• Proceedings of the Korean Society for Technology of Plasticity Conference
• /
• 1994.06a
• /
• pp.191-198
• /
• 1994

A practical example of the axisymmetric deep drawing process is simulated by the elastic-plastic finite element analysis using updated Lagrangian approach considering the large deformation. An approach is suggested to solve the problem of the ductile fracture that may encounter during the deep drawing process. The result can be applied to the design of the die for the axisymmetric deep drawing.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2000.04a
• /
• pp.721-723
• /
• 2000

Deep Drawing공정은 2차원 박판(Sheet Metal)에 그림이나 글자를 인쇄한 다음, 박판을 다이에 고정하고 펀치로 눌러서 3차원의 제품을 생산하는 소성가공의 한 방법이다. 그러므로, 2차원 평면인 박판에 어떻게 적절히 인쇄하여, 가공 후의 3차원 제품에 원하는 그림과 글자가 나타나게 할 수 있는지가 문제가 되고 있다. 본 논문에서는 Deep Drawing공정을 거쳐 완성된 제품을 측정한 후, 형상 역공학(Reverse Engineering) 기술을 이용하여 측정 데이터(Measured Points Data)를 입력으로 하는 매개변수 곡면 (Parametric Surface)을 만들고, Deep Drawing공정 전의 박판에 대한 매개변수 곡면을 만든 다음 두 곡면간의 매핑을 통해 위의 문제점을 해결하고자 한다.

• Transactions of Materials Processing
• /
• v.11 no.8
• /
• pp.676-681
• /
• 2002

The square cup deep drawing simulations for hydraulic counter pressure deep drawing are carried out by the finite element method and the formability factors which affect to the formability in case of that process are investigated. As a result, it is found that the thickness distributions keep the higher quality than that of the conventional deep drawing, and the maximum pressure increased the thickness at the die profile regions of blank. But friction coefficient decreased the thickness at the same regions.

• Proceedings of the Korean Geotechical Society Conference
• /
• 2002.10a
• /
• pp.85-95
• /
• 2002

A substantial portion of the cost of deep excavations in urban environments is devoted to prevent ground movements and their effects on adjacent buildings and utilites. Prediction of ground movements and assessment of the risk of damage to adjacent structures has become an essential part of the planning, design, and construction of a deep excavation project in the urban environments. This paper presents damage assessment techniques for buildings and utilities adjacent deep excavation, which can be readily used in practice.

• Journal of Advanced Information Technology and Convergence
• /
• v.8 no.2
• /
• pp.119-127
• /
• 2018

Training a very large deep neural network can be painfully slow and prone to overfitting. Many researches have done for overcoming the problem. In this paper, a combination of early stopping and ADAM based deep neural network was presented. This form of deep network is useful for handling the big data because it automatically stop the training before overfitting occurs. Also generalization ability is better than pure deep neural network model.

• Journal of the Korea Convergence Society
• /
• v.12 no.10
• /
• pp.55-61
• /
• 2021

In this paper, we proposed a DeepLabv3+ based encoder-decoder model utilizing an attention mechanism for precise semantic segmentation. The DeepLabv3+ is a semantic segmentation method based on deep learning and is mainly used in applications such as autonomous vehicles, and infrared image analysis. In the conventional DeepLabv3+, there is little use of the encoder's intermediate feature map in the decoder part, resulting in loss in restoration process. Such restoration loss causes a problem of reducing segmentation accuracy. Therefore, the proposed method firstly minimized the restoration loss by additionally using one intermediate feature map. Furthermore, we fused hierarchically from small feature map in order to effectively utilize this. Finally, we applied an attention mechanism to the decoder to maximize the decoder's ability to converge intermediate feature maps. We evaluated the proposed method on the Cityscapes dataset, which is commonly used for street scene image segmentation research. Experiment results showed that our proposed method improved segmentation results compared to the conventional DeepLabv3+. The proposed method can be used in applications that require high accuracy.

• Journal of Mechanical Science and Technology
• /
• v.15 no.4
• /
• pp.459-464
• /
• 2001

The effects of internal air-pressing on deep drawability are investigated in this study to increase the deep drawability of aluminum sheet. The conventional deep drawing process is limited to a certain limit drawing ratio(LDR) beyond which failure will occur. The intention of this work is to examine the possibilities of relaxing the above limitation through the deep drawing with internal air-pressing, aiming towards a process with an increased drawing ratio. The idea which may lead to this goal is the use of special punch that can exert high pressure on the internal surface of deforming sheet during the deep drawing process. Over the ranges of conditions investigated for Al-1050, the local strain concentration at punch nose radius area was decreased by internal air-pressing of punch, and the deep drawing with internal air-pressing was proved to be very effective process for obtaining higher LDR.

• Journal of the East Asian Society of Dietary Life
• /
• v.16 no.6
• /
• pp.717-723
• /
• 2006

Chemical characteristics of soybean oil after deep-frying with potato sticks (200 g, 10% w/w of soybean oil) were studied according to the 34 deep-frying times. After consecutive 34 deep-frying, total polyunsaturated FA contents was gradually decreased while the total saturated FA and trans FA were increased. Acid value and peroxide value were increased while iodine value decreased, respectively. The Hunter $L^{\ast}$ value decreased while each $a^{\ast}\;and \;b^{\ast}b$ value were gradually increased. Electronic nose equipped with 12 metal oxide sensors was used for the discrimination of odor pattern of frying oils against the times of deep-trying. The proportions of 1st and 2nd principal component analysis showed 75.97% and 21.23%, respectively. While 6 among total 12 sensors well responded to discrimination of odor in the repented frying oils, suggesting that the odor pattern of each oil after deep-frying would be discriminated against fresh soybean oil, especially after 14 times. From the results, electronic nose could differentiate the degree of quality deterioration of the repeated deep-frying oils.