• Journal of Korea Multimedia Society
• /
• v.22 no.5
• /
• pp.547-557
• /
• 2019

To extract the depth map from a single image, a number of CNN-based deep learning methods have been performed in recent research. In this study, the GAN structure of Pix2Pix is maintained. this model allows to converge well, because it has the structure of the generator and the discriminator. But the convolution in this model takes a long time to compute. So we change the convolution form in the generator to a depthwise convolution to improve the speed while preserving the result. Thus, the seven down-sizing convolutional hidden layers in the generator U-Net are changed to depthwise convolution. This type of convolution decreases the number of parameters, and also speeds up computation time. The proposed model shows similar depth map prediction results as in the case of the existing structure, and the computation time in case of a inference is decreased by 64%.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.9 no.1
• /
• pp.141-150
• /
• 2013

This study proposes technology using Active Shape Model to track the object separating it by depth-sensors. Unlike the common visual camera, the depth-sensor is not affected by the intensity of illumination, and therefore a more robust object can be extracted. The proposed algorithm removes the horizontal component from the information of the initial depth map and separates the object using the vertical component. In addition, it is also a more efficient morphology, and labeling to perform image correction and object extraction. By applying Active Shape Model to the information of an extracted object, it can track the object more robustly. Active Shape Model has a robust feature-to-object occlusion phenomenon. In comparison to visual camera-based object tracking algorithms, the proposed technology, using the existing depth of the sensor, is more efficient and robust at object tracking. Experimental results, show that the proposed ASM-based algorithm using depth sensor can robustly track objects in real-time.

• KCID journal
• /
• v.14 no.2
• /
• pp.225-235
• /
• 2007

The shape of streamflow hydrograph during the early period of spring is very much controlled by the area and depth of snow cover especially in mountainous area. When we simulate the streamfolw of a watershed snowmelt, we need some information for snow cover extent and depth distribution as parameters and input data in the hydrological models. The purpose of this study is to suggest an extraction method of snow cover area and snow depth distribution using Terra MODIS image. Snow cover extent for South Korea was extracted for the period of December 2000 and April 2006. For the snow cover area, the snow depth was interpolated using the snow depth data from 69 meteorological observation stations. With these data, it is necessary to run a hydrological model considering the snow-related data and compare the simulated streamflow with the observed data and check the applicability for the snowmelt simulation.

• Proceedings of the Costume Culture Conference
• /
• 2005.10a
• /
• pp.48-54
• /
• 2005

In order to mass-customize clothes, it is essential to take into account individual body shape using computerized 3D body models. This paper describes the development of an interactive body model that can be altered to match individual body perimeter, postures and depth for the purpose of computerized pattern making. Construction of the body model requires the extraction of necessary points, adjustment of coordinate points, linking of points by spline curves, control of section lengths and selectability of various hip types. Front to back depth of the model is adjusted by scaling ratio. We had a great result for controlling perimeter, posture and depth of body shapes. The results support the adaptability and potential usefulness of the posture and depth adjustable body model.

• Magazine of the Korean Society of Agricultural Engineers
• /
• v.37 no.5
• /
• pp.35-42
• /
• 1995

his paper is to develop a GIS application model (GISCELWAB) for the spatial simulation of surface runoff from a small watershed. The model was constituted by three submodels : The input data extraction model (GISINDATA) which prepares cell-based input data automatically for a given watershed, the cell water balance model (CELWAB) which calculates the water balance for a cell and simulates surface runoff of watershed simultaneously by the interaction of cells, and the output data management model (GISOUTDISP) which visualize the results of temporal and spatial variation of surface runoff. The input data extraction model was developed to solve the time-consuming problems for the input-data preparation of distributed hydrologic model. The input data for CELWAB can be obtained by extracting ASCII data from a vector map. The output data management model was developed to convert the storage depth and discharge of cells into grid map. This model enables to visualize the spatial formulation process of watershed storage depth and surface runoff wholly with time increment.

• Journal of Information Processing Systems
• /
• v.16 no.1
• /
• pp.155-170
• /
• 2020

Passive ranging is a critical part of machine vision measurement. Most of passive ranging methods based on machine vision use binocular technology which need strict hardware conditions and lack of universality. To measure the distance of an object placed on horizontal plane, we present a passive ranging method based on monocular vision system by smartphone. Experimental results show that given the same abscissas, the ordinatesis of the image points linearly related to their actual imaging angles. According to this principle, we first establish a depth extraction model by assuming a linear function and substituting the actual imaging angles and ordinates of the special conjugate points into the linear function. The vertical distance of the target object to the optical axis is then calculated according to imaging principle of camera, and the passive ranging can be derived by depth and vertical distance to the optical axis of target object. Experimental results show that ranging by this method has a higher accuracy compare with others based on binocular vision system. The mean relative error of the depth measurement is 0.937% when the distance is within 3 m. When it is 3-10 m, the mean relative error is 1.71%. Compared with other methods based on monocular vision system, the method does not need to calibrate before ranging and avoids the error caused by data fitting.

• Journal of Electrical Engineering and Technology
• /
• v.11 no.6
• /
• pp.1857-1862
• /
• 2016

Human activity recognition using depth information is an emerging and challenging technology in computer vision due to its considerable attention by many practical applications such as smart home/office system, personal health care and 3D video games. This paper presents a novel framework of 3D human body detection, tracking and recognition from depth video sequences using spatiotemporal features and modified HMM. To detect human silhouette, raw depth data is examined to extract human silhouette by considering spatial continuity and constraints of human motion information. While, frame differentiation is used to track human movements. Features extraction mechanism consists of spatial depth shape features and temporal joints features are used to improve classification performance. Both of these features are fused together to recognize different activities using the modified hidden Markov model (M-HMM). The proposed approach is evaluated on two challenging depth video datasets. Moreover, our system has significant abilities to handle subject's body parts rotation and body parts missing which provide major contributions in human activity recognition.

• The Journal of the Korea Contents Association
• /
• v.18 no.11
• /
• pp.447-454
• /
• 2018

This paper proposes a depth image generation algorithm of stereo images using a deep learning model composed of a CNN (convolutional neural network). The proposed algorithm consists of a feature extraction unit which extracts the main features of each parallax image and a depth learning unit which learns the parallax information using extracted features. First, the feature extraction unit extracts a feature map for each parallax image through the Xception module and the ASPP(Atrous spatial pyramid pooling) module, which are composed of 2D CNN layers. Then, the feature map for each parallax is accumulated in 3D form according to the time difference and the depth image is estimated after passing through the depth learning unit for learning the depth estimation weight through 3D CNN. The proposed algorithm estimates the depth of object region more accurately than other algorithms.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.1
• /
• pp.649-654
• /
• 2023

This paper proposes the design of a neural network structure search model using graph convolutional neural networks. Deep learning has a problem of not being able to verify whether the designed model has a structure with optimized performance due to the nature of learning as a black box. The neural network structure search model is composed of a recurrent neural network that creates a model and a convolutional neural network that is the generated network. Conventional neural network structure search models use recurrent neural networks, but in this paper, we propose GC-NAS, which uses graph convolutional neural networks instead of recurrent neural networks to create convolutional neural network models. The proposed GC-NAS uses the Layer Extraction Block to explore depth, and the Hyper Parameter Prediction Block to explore spatial and temporal information (hyper parameters) based on depth information in parallel. Therefore, since the depth information is reflected, the search area is wider, and the purpose of the search area of the model is clear by conducting a parallel search with depth information, so it is judged to be superior in theoretical structure compared to GC-NAS. GC-NAS is expected to solve the problem of the high-dimensional time axis and the range of spatial search of recurrent neural networks in the existing neural network structure search model through the graph convolutional neural network block and graph generation algorithm. In addition, we hope that the GC-NAS proposed in this paper will serve as an opportunity for active research on the application of graph convolutional neural networks to neural network structure search.

• Journal of KIISE:Computer Systems and Theory
• /
• v.31 no.1_2
• /
• pp.59-66
• /
• 2004

In this paper, a fast 3D model extraction algorithm with an enhanced PBIL of preserving depth consistency is proposed for the extraction of 3D depth information from 2D images. Evolutionary computation algorithms are efficient search methods based on natural selection and population genetics. 2D disparity maps acquired by conventional matching algorithms do not match well with the original image profile in disparity edge regions because of the loss of fine and precise information in the regions. Therefore, in order to decrease the imprecision of disparity values and increase the quality of matching, a compact genetic algorithm is adapted for matching environments, and the adaptive window, which is controlled by the complexity of neighbor disparities in an abrupt disparity point is used. As the result, the proposed algorithm showed more correct and precise disparities were obtained than those by conventional matching methods with relaxation scheme.