Modern computer systems usually have special hardware for operations used in deep learning workload even edge computing environment. Non-volatile memories (NVMs) have been considered for alternative memory storage because they consume little static energy and occupy small area. However, there is a problem for NVMs to be directly adopted. An NVM cell has limited write endurance, so that the lifetime of NVM-based memory system is much shorter than that of conventional memory system. To overcome this problem for the deep learning system, this paper proposes a novel method to extend the lifetime based on the analysis of the deep learning workloads. If an incoming block has more than a predefined number of frequently used values, the cacheline is defined as write friendly block. During the victim selection, the cacheline has lower possibility to be chosen as victim. The experimental results show that the lifetime is increased by about 50% and energy consumption is decreased by 3% with a little performance hurt.

Object detection technology is one of the main research topics in the field of computer vision and has established itself as an essential base technology for implementing various vision systems. Recent DNN (Deep Neural Networks)-based algorithms achieve much higher recognition accuracy than traditional algorithms. However, it is well-known that the DNN model inference operation requires a relatively high computational power. In this paper, we analyze the inference time complexity of the state-of-the-art object detection architecture Yolov7 in various environments. Specifically, we compare and analyze the time complexity of four types of the Yolov7 model, YOLOv7-tiny, YOLOv7, YOLOv7-X, and YOLOv7-E6 when performing inference operations using CPU and GPU. Furthermore, we analyze the time complexity variation when inferring the same models using the Pytorch framework and the Onnxruntime engine.

In this paper, a method for measuring image pattern widths is proposed using gaussian interpolation, in order to improve inconsistent results coming from the different directions in image patterns. The performance of our method is evaluated using image patterns with 9 directions, and compared with previous methods. It is confirmed that the proposed method gives accurate and consistent width results regardless of pattern directions.

A variety of techniques have been proposed in the literature for depth improvement in depth from focus method. Unfortunately, these techniques over-smooth the depth maps over the regions of depth discontinuities. In this paper, we propose a robust technique for improving the depth map by employing a nonconvex smoothness function that preserves the depth edges. In addition, the proposed technique exploits the mutual structures between the depth map and a guidance map. This guidance map is designed by taking the mean of image intensities in the image sequence. The depth map is updated iteratively till the nonconvex objective function converges. Experiments performed on real complex image sequences revealed the effectiveness of the proposed technique.

Metal oxide semiconductor (MOS) based on spray-pyrolysis deposition technique has attracted large attention due to simple and low-cost processibility while preserving their intrinsic optical and electrical characteristics. However, their high process temperature limits practical applications. Here, we demonstrated the nc-ZnO/ZnO field-effect transistors (FETs) via spray-pyrolysis as incorporating ZnO nanocrystalline nanoparticles into typical ZnO precursor. The nc-ZnO/ZnO FETs exhibit good quality of electrical properties. Our experiments reveal that nc-ZnO in active layer enhance electrical characteristics.

EtherCAT is an Ethernet-based fieldbus system standardized in IEC 61158 and SEMI, and widely used in the fields of factory automation, semiconductor equipment and robotics. In this paper, an EtherCAT master is implemented on an embedded board with Arm based 64-bit quad-core processor and its jitter performance is evaluated at the output of the network interface to include all the effects of the entire system in the results. For the EtherCAT master system, an open source EtherCAT master stack, Simple Open EtherCAT Master (SOEM), is installed on PREEMPT_RT patched Linux operating system for real-time operation. The results show that the jitter performance is comparable to that of Xenomai-based master and the EtherCAT master with two master instances has similar jitter performance to the EtherCAT master with one master instance.

A solar power generation system uses a solar module that collects solar radiation energy, a connecting board that collects DC power generated from the solar module, and a diode to prevent reverse current from flowing from an inverter to the solar module. The existing photovoltaic connection module consists of only fuses and diodes for reverse polarity and overcurrent blocking, and does not have fire diagnosis, prevention, and suppression functions in the event of a fire. To solve this problem, this paper presents a method to monitor the internal state of the photovoltaic connection module using several sensors and to prevent and extinguish a fire using solenoid valves and fire extinguishing agents when a fire is detected. Through the experiment, it was confirmed that the proposed method normally suppresses the fire in event of a fire.

The light extraction technology for improving the light efficiency of OLEDs is the core technology for extracting the light inside the OLEDs to the outside. This study demonstrates a simple method to generate random nanostructures (RNSs) containing high refractive index nanoparticles to improve light extraction and viewing angle characteristics. A simple dry low-temperature process makes the nanostructured scattering layer on the polymer resin widely used in the industry. The scattering layer has the shape of randomly distributed nanorods. To control optical properties, we focused on changing the shape and density of RNSs and adjusting the concentration of high refractive index nanoparticles. As a result, the film of the present invention exhibits a perpendicular transmittance of 85% at a wavelength of 550 nm. This film was used as a scattering layer to reduce substrate mode loss and improve EL efficiency in OLEDs.

The fourth industrial revolution, internet of things, and the expansion of online web services have increased an exponential growth and deployment in the number of cloud data centers (CDC). The cloud is emerging as new paradigm for delivering the Internet-based computing services. Due to the dynamic and non-linear workload and availability of the resources is a critical problem for efficient workload and resource management. In this paper, we propose the particle swarm optimization (PSO) based gated recurrent unit (GRU) neural network for efficient prediction the future value of the CPU and memory usage in the cloud data centers. We investigate the hyper-parameters of the GRU for better model to effectively predict the cloud resources. We use the Google Cluster traces to evaluate the aforementioned PSO-GRU prediction. The experimental shows the effectiveness of the proposed algorithm.

In order to process a vast amount of data, there is demand for a new system with higher processing speed and lower energy consumption. To prevent 'memory wall' in von Neumann architecture, RRAM, which is a neuromorphic device, has been researched. In this paper, we summarize the features of RRAM and propose the device structure for characteristic improvement. RRAM operates as a synapse device using a change of resistance. In general, the resistance characteristics of RRAM are nonlinear and random. As synapse device, linearity and uniformity improvement of RRAM is important to improve learning recognition rate because high linearity and uniformity characteristics can achieve high recognition rate. There are many method, such as TEL, barrier layer, NC, high oxidation properties, to improve linearity and uniformity. We proposed a new device structure of TiN/Al doped TaO_x/AlO_x/Pt that will achieve high recognition rate. Also, with simulation, we prove that the improved properties show a high learning recognition rate.

The plasma resistance of multi-component glasses containing La, Gd, Ti, Zn, Y, Zr, Nb, and Ta was analyzed in this study. The plasma etching was performed via inductively coupled plasma-reactive ion etching (ICP-RIE) using CF₄/O₂/Ar mixed gas. After the reaction, the glass with a low fluoride sublimation temperature and high content of P, Si, and Ti elements showed a high etching rate. On the other hand, the glass containing a high fluoride sublimation temperature component such as Ca, La, Gd, Y, and Zr exhibited high plasma resistance because the etch rate was lower than that of sapphire. Glass with low plasma resistance increased surface roughness after etching or nanoholes were formed on the surface, but glass with high plasma resistance showed little change in surface microstructure. Thus, the results of this study demonstrate the potential for the development of plasma-resistant glasses (PRGs) with other compositions besides alumino-silicate glasses, which are conventionally referred to as plasma-resistant glasses.

In this study, we produced defect detection models using parallel convolution neural networks. If convolution neural networks are constructed parallel type, the model's detection accuracy will increase and detection time will decrease. We produced parallel-type defect detection models using 4 types of convolutional algorithms. The performance of models was evaluated using evaluation indicators. The model's performance is detection accuracy and detection time. We compared the performance of each parallel model. The detection accuracy of the model using AlexNet is 97 % and the detection time is 0.3 seconds. We confirmed that when AlexNet algorithm is constructed parallel type, the model has the highest performance.

The application of cloud computing to many industrial domains is rapidly increasing, and domestic and foreign cloud service providers are actively conducting business. In the domestic cloud market, it is necessary to establish an ecosystem with partner operators that work closely with private cloud service providers. In this paper, to create such an environment, we propose a framework that can evaluate the capabilities of partners required for cloud service providers to establish specific business strategies. The framework proposed in this study establishes criteria for evaluating partners' competencies and applies a decision-making model such as fuzzy AHP for evaluation. Eventually this will help not only to expand the domestic cloud market but also to strengthen the competitiveness of domestic cloud partners through the growth of the domestic cloud market.

In this study, the plasma corrosion resistance and the change in the number of contamination particles generated using the plasma etching process and cleaning process of coating parts for semiconductor plasma etching equipment were investigated. As the coating method, atmospheric plasma spray (APS) was used, and the powder materials were Y₂O₃ and Y₃Al₅O₁₂ (YAG). There was a clear difference in the densities of the coatings due to the difference in solubility due to the melting point of the powdered material. As a plasma environment, a mixed gas of CF₄, O₂, and Ar was used, and the etching process was performed at 200 W for 60 min. After the plasma etching process, a fluorinated film was formed on the surface, and it was confirmed that the plasma resistance was lowered and contaminant particles were generated. We performed a surface cleaning process using piranha solution(H₂SO₄(3):H₂O₂(1)) to remove the defect-causing surface fluorinated film. APS-Y₂O₃ and APS-YAG coatings commonly increased the number of defects (pores, cracks) on the coating surface by plasma etching and cleaning processes. As a result, it was confirmed that the generation of contamination particles increased and the breakdown voltage decreased. In particular, in the case of APS-YAG under the same cleaning process conditions, some of the fluorinated film remained and surface defects increased, which accelerated the increase in the number of contamination particles after cleaning. These results suggest that contaminating particles and the breakdown voltage that causes defects in semiconductor devices can be controlled through the optimization of the APS coating process and cleaning process.

In this paper, the optimal design and circuit simulation verification results of an LLC resonant half-bridge dc-dc converter using a steady-state model with internal loss resistance are reported. Above all, the input/output voltage gain and frequency characteristic equations in the steady-state were derived by reflecting the internal loss resistance in the equivalent circuit. Based on the results, an LLC resonant half-bridge dc-dc converter with an input voltage of 360-420V, an output voltage of 54V, and a maximum power of 3kW was designed, and to verify the design, the PSIM circuit simulation was executed to compare and analyze the result. In particular, the operating range of the converter could be drawn from the frequency characteristic graph of the voltage gain, and when the converter was operated under light and maximum load conditions, it was confirmed that similar results were obtained by comparing simulation results and calculation results in the switching frequency characteristic graph. In addition, the change of the switching frequency with respect to the load current at each input voltage was compared with the calculated value and the simulation result. As a result, it was possible to confirm the usefulness of the analysis result reflecting the internal loss resistance proposed in this paper and the process of the optimal design.

Bokeh effect is a stylistic technique that can produce blurring the background of photos. This paper implements to produce a bokeh effect with a single image by post processing. Generating depth map is a key process of bokeh effect, and depth map is an image that contains information relating to the distance of the surfaces of scene objects from a viewpoint. First, this work presents algorithms to determine the depth map from a single input image. Then, we obtain a sparse defocus map with gradient ratio from input image and blurred image. Defocus map is obtained by propagating threshold values from edges using matting Laplacian. Finally, we obtain the blurred image on foreground and background segmentation with bokeh effect achieved. With the experimental results, an efficient image processing method with bokeh effect applied using a single image is presented.

Plastics are classified into 7 types such as PET (PETE), HDPE, PVC, LDPE, PP, PS, and Other for separation and recycling. Recently, large corporations advocating ESG management are replacing them with bioplastics. Incineration and landfill of disposal of plastic waste are responsible for air pollution and destruction of the ecosystem. Because it is not easy to accurately classify plastic materials with the naked eye, automated system-based screening studies using various sensor technologies and AI-based software technologies have been conducted. In this paper, NIR scanning devices considering the NIR wavelength characteristics that appear differently for each plastic material and a system that can identify the type of plastic by learning the NIR spectrum data collected through it. The accuracy of plastic material identification was evaluated through a decision tree-based SVM model for multiclass classification on NIR spectral datasets for 8 types of plastic samples including biodegradable plastic.

This paper presents a new blockchain based donation platform combined with a type of gaming application. The proposed donation platform is similar to existing play-to-earn (P2E), but it has different characteristics such as non-liquidated token and direct transfer to a celebrity donation organization. Moreover, it has advantages like a robustness for counterfeiting and falsification as well as enhanced donating experience.

Mist-CVD is known to have advantages of low cost and high productivity compared to ALD and PECVD methods. It is capable of reacting to the substrate by misting an aqueous solution using ultrasonic waves under vacuum-free conditions of atmospheric pressure. In particular, Ga₂O₃ is regarded as advanced power semiconductor material because of its high quality of transmittance, and excellent electrical conductivity through N-type doping. In this study, Computational Fluid Dynamics were used to predict the uniformity of the thin film on a large-area substrate. And also the deposition pattern and uniformity were analyzed using the flow velocity and particle tracking method. The uniformity was confirmed by quantifying the deposition cross section with an FIB-SEM, and the consistency of the uniformity prediction was secured through the analysis of the CFD distribution. With the analysis and experimental results, the match rate of deposition area was 80.14% and the match rate of deposition thickness was 55.32%. As the experimental and analysis results were consistent, it was confirmed that it is possible to predict the deposition thickness uniformity of Mist-CVD.

For the realization of higher reliable transition metal dichalcogenide layer, molybdenum disulfide was formed on sapphire substrate by direct current sputtering and subsequent rapid thermal annealing process. Unlike RF sputtered MoS₂ thin films, DC sputtered showed no irregular holes and protrusions after annealing process from scanning electron microscope images. From atomic force microscope results, it was possible to investigate that surface roughness of MoS₂ thin films were more dependent on DC sputtering power then annealing temperature. On the other hand, the Raman scattering spectra showed the dependency of significant E¹_2g and A¹_g peaks on annealing temperatures.

Non-playable game character plays an important role in improving the concentration of the game and the interest of the user, and recently implementation of NPC with reinforcement learning has been in the spotlight. In this paper, we estimate an AI target tracking method via reinforcement learning, and implement an AI-based tracking agency of specific target object with avoiding traps through Unity ML-Agents Toolkit. The implementation is built in Unity game engine, and simulations are conducted through a number of experiments. The experimental results show that outstanding performance of the tracking target with avoiding traps is shown with good enough results.

The stabilized all-solid-state battery structure indicate a fundamental alternative to the development of next-generation energy storage devices. Existing liquid electrolyte structures severely limit battery stability, creating safety concerns due to the growth of Li dendrites during rapid charge/discharge cycles. In this study, a low-dimensional graphene quantum dot layer structure was applied to demonstrate stable operating characteristics based on Li+ ion conductivity and excellent electrochemical performance. Transmission electron microscopy analysis was performed to elucidate the microstructure at the interface. The low-dimensional structure of GQD-based solid electrolytes has provided an important strategy for stable scalable solid-state lithium battery applications at room temperature. This study indicates that the low-dimensional carbon structure of Li-GQDs can be an effective approach for the stabilization of solid-state Li matrix architectures.

Top emission organic light-emitting diode (TEOLED) is commonly used because of high efficiency and good color purity than bottom - emission organic light-emitting device (BEOLED). Unlike BEOLED, TEOLED contain semitransparent metal cathode and capping layer. Because there are many characteristics to consider just simple thickness change, optimizing organic thickness of TEOLED for microcavity is difficult. So, in this study, we optimized Device capping layer at unoptimized micro-cavity structure TEOLED device. And we compare only capping layer with unoptimized microcavity structure can overcome optimized micro-cavity structure device. We used previous our optimized micro-cavity structure to compare each other. As a result, it has been found that the efficiency can be obtained almost the same or higher only capping layer, which is stacked on top of the device and controls only the thickness and refractive index, without complicated structural calculations. This means that higher efficiencies can be obtained more easily in laboratories with limited organic materials or when optimizing new structures etc.

The current display technology tends to be highly integrated with high resolution, the element size is gradually downsized, and the structure becomes complicated. Inductively coupled plasma (ICP) dry etcher of various types of etching equipment is a structure that places a large multi-divisional antenna source on the top lid, passes current to the Antenna, and generates plasma using the induced magnetic field generated at this time. However, in the case of a device of a large area size, a support that can withstand a load structurally is necessary, and when these support portions are applied, arrangement of antenna becomes difficult, which causes reduction in uniformity. As described above, the development of antenna source of a large area having a uniform plasma density on the whole surface is difficult to restrict hardware (H/W). As a solution to this problem, we confirmed the change in uniformity of plasma by applying two kinds of specific shape faraday shield(FICP) to the lower part of the large area upper lid antenna of 6 and 8th more than that generation size. In this thesis, we verify the faraday shield effect which can improve plasma uniformity control of ICP dry etcher equipment applied to medium and large displays.

An anti-reflective coating is used to improve the performance of the solar cell. The anti-reflective coating changes the value of the short-circuit current about the thickness. However, the current-voltage characteristics about the anti-reflective coating are difficult to calculate without simulation tool. In this paper, a modeling technique to determine the short-circuit current value and the current-voltage characteristics in accordance with the thickness is proposed. In addition, artificial neural network is used to predict the short-circuit current with the dependence of temperature and thickness. Simulation results incorporating the artificial neural network model are obtained using MATLAB/Simulink and show the current-voltage characteristic according to the thickness of the anti-reflective coating.

The current-voltage modeling plays an important role in characterizing photovoltaic systems. A solar cell has a nonlinear characteristic with various parameters influenced by the external environments such as the irradiance and the temperature. In order to accurately predict current-voltage characteristics at low irradiance, the artificial neural networks are applied to effectively quantify nonlinear behaviors. In this paper, a multi-layer perceptron scheme that can make accurate predictions is employed to learn complex formulas for large amounts of continuous data. The simulated results of artificial neural networks model show the accuracy improvement by using MATLAB/Simulink.