• 수산해양기술연구
• /
• 제46권4호
• /
• pp.376-391
• /
• 2010

To investigate seasonal variation and species composition by depth layers in the deep sea ecosystem of the East Sea of Korea, bottom trawl survey was conducted at 4 depth layers during spring and autumn from 2007 to 2009. A total of 47 species were collected and were composed of 23 fish species, 9 crustacea, 6 cephalopoda and 9 gastropoda. The main dominant species at each depth layers were Chionoecetes opilio in 300m, Berryteuthis magister in 500m, Chionoecetes japonicus in 700m and 900m. In spring, richness indices (R) showed low value of 2.01 in 500m depth, and high value of 2.16 in 300m depth. Diversity indices (H') showed low value of 1.53 in 300m depth, and high value of 2.09 in 700m depth. Dominance indices (D) showed low value of 0.15 in 700m depth, and high value of 0.31 in 300m depth. In Autumn, richness indices showed low value of 1.48 in 900m depth, and high value of 2.69 in 300m depth. Diversity' indices (H') showed low value of 1.13 in 300m depth, and high value of 2.23 in 700m depth. Dominance indices (D) showed low value of 0.14 in 700m depth, and high value of 0.54 in 300m depth. In spring, similarity analysis in each depth layers showed the difference between 900m and othe depth layer, on the contrary 500m and 700m showed the similarity. In autumn, similarity analyssis in each depth layers showed the difference between 700m and other depth layers, on the contrary 300m and 500m showed the similarity.

• Smart Structures and Systems
• /
• 제16권3호
• /
• pp.497-519
• /
• 2015

This study discusses the use of Adaptive-Network-Based-Fuzzy-Inference-System (ANFIS) in predicting the shear strength of reinforced-concrete deep beams. 139 experimental data have been collected from renowned publications on simply supported high strength concrete deep beams. The results show that the ANFIS has strong potential as a feasible tool for predicting the shear strength of deep beams within the range of the considered input parameters. ANFIS's results are highly accurate, precise and therefore, more satisfactory. Based on the Sensitivity analysis, the shear span to depth ratio (a/d) and concrete cylinder strength ($f_c^{\prime}$) have major influence on the shear strength prediction of deep beams. The parametric study confirms the increase in shear strength of deep beams with an equal increase in the concrete strength and decrease in the shear span to-depth-ratio.

• 방송공학회논문지
• /
• 제23권5호
• /
• pp.614-621
• /
• 2018

본 논문에서는 generative adversarial network (GAN)을 이용한 비감독 학습을 통해 깊이 카메라로 깊이 영상을 취득할 때 발생한 손실된 부분을 복원하는 기법을 제안한다. 제안하는 기법은 3D morphable model convolutional neural network (3DMM CNN)와 large-scale CelebFaces Attribute (CelebA) 데이터 셋 그리고 FaceWarehouse 데이터 셋을 이용하여 학습용 얼굴 깊이 영상을 생성하고 deep convolutional GAN (DCGAN)의 생성자(generator)와 Wasserstein distance를 손실함수로 적용한 구별자(discriminator)를 미니맥스 게임기법을 통해 학습시킨다. 이후 학습된 생성자와 손실 부분을 복원해주기 위한 새로운 손실함수를 이용하여 또 다른 학습을 통해 최종적으로 깊이 카메라로 취득된 얼굴 깊이 영상의 손실 부분을 복원한다.

• 한국농공학회지
• /
• 제44권2호
• /
• pp.67-74
• /
• 2002

The purpose of this study was to investigate the effects of ponding depth treatment on water balance in paddy fields. Three ponding depth treatments, very shallow, shallow, and deep were used. The experimental plots were three 80m $\times$ 8m rectangular plots. Daily values of rainfall amount, ponding depth, irrigation water, drainage water, evapotranspiration, and infiltration were measured in the field. The ponding depth was continuously observed by water level logger during the growing season. The ET was measured by 1-m diameter PVC lysimeters. Irrigation water volume was measured by 75 mm pipe flowmeters and the drainage water volume by 75 mm pipe flowmeters and a recording Parshall flume. The results showed that irrigation water depths were 688.9 mm, 513.6 mm, and 624.4 mm in very shallow, shallow, and deep ponding, respectively. The effective rainfall amounts (effective ratio) were 243.7 mm(48.8%), 344.6 mm(68.9%), and 272.9 mm(54.6%) in very shallow, shallow, and deep ponding, respectively. The three treatments did not show any statistical difference in growth and yields. But the shallow depth treatment showed the largest yield.

• Journal of Electrical Engineering and Technology
• /
• 제1권2호
• /
• pp.233-236
• /
• 2006

A 3-D numerical simulation of a buried-channel CCD (Charge Coupled Device) with a deep depletion has been performed to investigate its electrical and physical behaviors. Results are presented for a deep depletion CCD (EEV CCD12; JET-X CCD) fabricated on a high-resistivity $(1.5k\Omega-cm)\;65{\mu}m$ thick epi-layer, on a $550{\mu}m$ thick p+ substrate, which is optimized for X-ray detection. Accurate predictions of the Potential minimum and barrier height of a CCD Pixel as a function of mobile electrons are found to give good charge transfer. The depletion depth approximation as a function of gate and substrate bias voltage provided average errors of less than 6%, compared with the results estimated from X-ray detection efficiency measurements. The result obtained from the transient simulation of signal charge movement is also presented based on 3-Dimensional analysis.

• 로봇학회논문지
• /
• 제14권2호
• /
• pp.114-121
• /
• 2019

This paper presents a new benchmark system for visual odometry (VO) and monocular depth estimation (MDE). As deep learning has become a key technology in computer vision, many researchers are trying to apply deep learning to VO and MDE. Just a couple of years ago, they were independently studied in a supervised way, but now they are coupled and trained together in an unsupervised way. However, before designing fancy models and losses, we have to customize datasets to use them for training and testing. After training, the model has to be compared with the existing models, which is also a huge burden. The benchmark provides input dataset ready-to-use for VO and MDE research in 'tfrecords' format and output dataset that includes model checkpoints and inference results of the existing models. It also provides various tools for data formatting, training, and evaluation. In the experiments, the exsiting models were evaluated to verify their performances presented in the corresponding papers and we found that the evaluation result is inferior to the presented performances.

• 한국염색가공학회지
• /
• 제26권3호
• /
• pp.187-194
• /
• 2014

To improve the deep coloring effect of PET fabrics, the alkali treated and black dyed PET fabrics were treated with 2 kinds of low refractive compounds such as acrylic resin and silicone resin. The color depth effect of treated PET fabrics was evaluated as lightness(L) change by UV-visible spectrophotometer. As the weight loss of PET fiber treated with alkali increased, the color depth of PET fabrics increased. Lightness(L) of PET fabrics treated with deep coloring agent was lower than that of untreated PET fabrics. The optimum concentration of treated PET with deep coloring agent was 4% o.w.s. The deep coloring effect of PET fabrics treated with silicone resin was higher than one treated with acrylic resin. PET fabrics treated with silicone resin only might be more appropriate process than PET fabrics treated with acrylic and silicone resin for giving deep coloring effect for polyester fabrics.

• International journal of advanced smart convergence
• /
• 제8권4호
• /
• pp.75-81
• /
• 2019

Though recent advancement of deep learning methods have provided satisfactory results from large data domain, somehow yield poor performance on few-shot classification tasks. In order to train a model with strong performance, i.e. deep convolutional neural network, it depends heavily on huge dataset and the labeled classes of the dataset can be extremely humongous. The cost of human annotation and scarcity of the data among the classes have drastically limited the capability of current image classification model. On the contrary, humans are excellent in terms of learning or recognizing new unseen classes with merely small set of labeled examples. Few-shot learning aims to train a classification model with limited labeled samples to recognize new classes that have neverseen during training process. In this paper, we increase the backbone depth of the embedding network in orderto learn the variation between the intra-class. By increasing the network depth of the embedding module, we are able to achieve competitive performance due to the minimized intra-class variation.

• 소성∙가공
• /
• 제4권4호
• /
• pp.345-352
• /
• 1995

Warm deep drawing for optimum forming conditions to give the maximum drawing depth is investigated and compared with the results from experiments performed at room temperature. Experiments which draw square cups of STS 304 stainless steel sheet under the constant lubrication condition of teflon film are made both in a crank and hydraulic press for two kinds of specimens. The maximum drawing depth at warm forming condition reaches 1.4 times the drawing depth at room temperature in a crank press, whereas 1.6 times in a hydraulic press, and also more uniform distribution of thickness in case of warm deep drawn cup is observed. The effects of other factors on formability, such as forming temperature, speed of press and cooling of punch are examined and discussed.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제11권2호
• /
• pp.1118-1133
• /
• 2017

Computer vision-based human activity recognition (HAR) has become very famous these days due to its applications in various fields such as smart home healthcare for elderly people. A video-based activity recognition system basically has many goals such as to react based on people's behavior that allows the systems to proactively assist them with their tasks. A novel approach is proposed in this work for depth video based human activity recognition using joint-based motion features of depth body shapes and Deep Belief Network (DBN). From depth video, different body parts of human activities are segmented first by means of a trained random forest. The motion features representing the magnitude and direction of each joint in next frame are extracted. Finally, the features are applied for training a DBN to be used for recognition later. The proposed HAR approach showed superior performance over conventional approaches on private and public datasets, indicating a prominent approach for practical applications in smartly controlled environments.