• Journal of Broadcast Engineering
• /
• v.23 no.5
• /
• pp.614-621
• /
• 2018

This paper proposes a method of restoring corrupted depth image captured by depth camera through unsupervised learning using generative adversarial network (GAN). The proposed method generates restored face depth images using 3D morphable model convolutional neural network (3DMM CNN) with large-scale CelebFaces Attribute (CelebA) and FaceWarehouse dataset for training deep convolutional generative adversarial network (DCGAN). The generator and discriminator equip with Wasserstein distance for loss function by utilizing minimax game. Then the DCGAN restore the loss of captured facial depth images by performing another learning procedure using trained generator and new loss function.

• The Journal of the Korea Contents Association
• /
• v.9 no.7
• /
• pp.57-66
• /
• 2009

Zigbee is a universal communication standard used in USN and is utilized in various applications. Zigbee protocol provides an address within a single PAN network, and at this time, it uses DAA. This is a method that divides a 16-bit address area into blocks with a fixed size according to the depth to assign one to each node. However, this method is limited because it has to assign addresses in 16 bits. As the depth increases, the number of nodes also increases exponentially to the maximal number of routers provided to each depth. Therefore, it is difficult to construct a huge network with numerous routers and large depth as in the places which are wide or have many shadow areas. Besides, since all the operations are performed in a single PAN network, it is hard to make several PANs into a single network. This article suggests new addressing and routing methods that can construct several PAN networks into a single network and combine broad area with less limitation in the number of routers and depth by extending the Zigbee-based network. Moreover, this paper has tested its performance and has verified its usability through substantive tests.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.19 no.8
• /
• pp.1907-1914
• /
• 1995

An on-line measurement of the workpiece hardened depth in laser surface hardening processes is very much difficult to achieve, since the hardening process occurs in depth wise direction. In this paper, the hardened depth is estimated using a multilayered neural network. Input data of the neural network are the surface temperatures at arbitrary chosen five surface points, laser power and traveling speed of laser beam torch. To simulate the actual hardening process, a finite difference method(FDM) is used to model the process. Since this model yields the calculation results of the temperature distribution around the workpiece volume in the vicinity of the laser torch, this model is used to obtain the network's training data and laser to evaluate the performance of the neural network estimator. The simulation results show that the proposed scheme can be used to estimate the hardened depth with reasonable accuracy.

• 제어로봇시스템학회:학술대회논문집
• /
• 1993.10a
• /
• pp.212-217
• /
• 1993

In this paper, the hardening depth in Laser surface hardening process is estimated using a multilayered neural network. Input data of the neural network are surface temperature of five points, power and travelling speed of Laser beam. A FDM(finite difference method) is used for modeling the Laser surface hardening process. This model is used to obtain the network's training data sample and to evaluate the performance of the neural network estimator. The simulational results showed that the proposed scheme can be used to estimate the hardening depth on real time.

• Journal of Broadcast Engineering
• /
• v.23 no.1
• /
• pp.3-10
• /
• 2018

Depth map has been utilized in a varity of fields. Recently research on generating depth map by artificial neural network (ANN) has gained much interest. This paper validates the feasibility of implementing the ready-made depth map generation by convolutional neural network (CNN). First, for a given image, a depth map is generated by the weighted average of a saliency map as well as a motion history image. Then CNN network is trained by test images and depth maps. The objective and subjective experiments are performed on the CNN and showed that the CNN can replace the ready-made depth generation method.

• 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 Information Processing Systems
• /
• v.19 no.4
• /
• pp.483-497
• /
• 2023

Depth estimation is one of the most complicated and difficult problems to deal with in the light field. In this paper, a compound attention convolutional neural network (CAttNet) is proposed to extract depth maps from light field images. To make more effective use of the sub-aperture images (SAIs) of light field and reduce the redundancy in SAIs, we use a compound attention mechanism to weigh the channel and space of the feature map after extracting the primary features, so it can more efficiently select the required view and the important area within the view. We modified various layers of feature extraction to make it more efficient and useful to extract features without adding parameters. By exploring the characteristics of light field, we increased the network depth and optimized the network structure to reduce the adverse impact of this change. CAttNet can efficiently utilize different SAIs correlations and features to generate a high-quality light field depth map. The experimental results show that CAttNet has advantages in both accuracy and time.

• Journal of Navigation and Port Research
• /
• v.28 no.7
• /
• pp.629-634
• /
• 2004

Continuous depth data are often required in applications of both onboard systems and maritime simulation. But data available are usually discrete and irregularly distributed. Based on the neuron network technique, methods of interpolation to the charted depth are suggested in this paper. Two algorithms based on Levenberg-Marquardt back-propaganda and radial-basis function networks are investigated respectively. A dynamic neuron network system is developed which satisfies both real time and mass processing applications. Using hyperbolic paraboloid and typical chart area, effectiveness of the algorithms is tested and error analysis presented. Special process in practical applications such as partition of lager areas, normalization and selection of depth contour data are also illustrated.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2004.08a
• /
• pp.37-44
• /
• 2004

Continuous depth data are often required in applications of both onboard systems and maritime simulation. But data available are usually discrete and irregularly distributed. Based on the neuron network technique, methods of interpolation to the charted depth are suggested in this paper. Two algorithms based on Levenberg-Marquardt back-propaganda and radial-basis function networks are investigated respectively. A dynamic neuron network system is developed which satisfies both real time and mass processing applications. Using hyperbolic paraboloid and typical chart area, effectiveness of the algorithms is tested and error analysis presented. Special process in practical applications such as partition of lager areas, normalization and selection of depth contour data are also illustrated.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.34 no.12B
• /
• pp.1475-1482
• /
• 2009

In ZigBee network, using energy efficiently is necessary because ZigBee node works by battery. To use energy efficiently, it is one of the way to reduce unnecessary network traffic. In this paper, it presents efficient shortcut routing algorithm using depth of destination node in ZigBee network. In traditional tree routing, each node transfers data only to its own parent or child node, which is inefficient way. Efficient shortcut routing algorithm is also based on tree routing. However, we suggests the algorithm with using neighbor table and depth of destination that is able to transfer data to other neighbor node, not only to parent or child node. It minimizes coordinator bottleneck state and unnecessary intermediate routing path which happens in traditional tree routing.