• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.3
• /
• pp.91-100
• /
• 2020

Unlike the existing Visual Question Answering(VQA) problems, the new Visual Commonsense Reasoning(VCR) problems require deep common sense reasoning for answering questions: recognizing specific relationship between two objects in the image, presenting the rationale of the answer. In this paper, we propose a novel deep neural network model, KG_VCR, for VCR problems. In addition to make use of visual relations and contextual information between objects extracted from input data (images, natural language questions, and response lists), the KG_VCR also utilizes commonsense knowledge embedding extracted from an external knowledge base called ConceptNet. Specifically the proposed model employs a Graph Convolutional Neural Network(GCN) module to obtain commonsense knowledge embedding from the retrieved ConceptNet knowledge graph. By conducting a series of experiments with the VCR benchmark dataset, we show that the proposed KG_VCR model outperforms both the state of the art(SOTA) VQA model and the R2C VCR model.

• Nuclear Engineering and Technology
• /
• v.54 no.12
• /
• pp.4751-4758
• /
• 2022

Two-phase flow may almost exist in every branch of the energy industry. For the corresponding engineering design, it is very essential and crucial to monitor flow patterns and their transitions accurately. With the high-speed development and success of deep learning based on convolutional neural network (CNN), the study of flow pattern identification recently almost focused on this methodology. Additionally, the photographing technique has attractive implementation features as well, since it is normally considerably less expensive than other techniques. The development of such a two-phase flow pattern online monitoring system is the objective of this work, which seldom studied before. The ongoing preliminary engineering design (including hardware and software) of the system are introduced. The flow pattern identification method based on CNNs and transfer learning was discussed in detail. Several potential CNN candidates such as ALexNet, VggNet16 and ResNets were introduced and compared with each other based on a flow pattern dataset. According to the results, ResNet50 is the most promising CNN network for the system owing to its high precision, fast classification and strong robustness. This work can be a reference for the online monitoring system design in the energy system.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.5
• /
• pp.169-176
• /
• 2020

Recently, deep neural networks (DNNs) have brought revolutions to many mobile and embedded devices by providing human-level machine intelligence for various applications. However, high inference accuracy of such DNNs comes at high computational costs, and, hence, there have been significant efforts to reduce computational overheads of DNNs either by compressing off-the-shelf models or by designing a new small footprint DNN architecture tailored to resource constrained devices. One notable recent paradigm in designing small footprint DNN models is sharing parameters in several layers. However, in previous approaches, the parameter-sharing techniques have been applied to large deep networks, such as ResNet, that are known to have high redundancy. In this paper, we propose a parameter-sharing method for already parameter-efficient small networks such as ShuffleNetV2. In our approach, small templates are combined with small layer-specific parameters to generate weights. Our experiment results on ImageNet and CIFAR100 datasets show that our approach can reduce the size of parameters by 15%-35% of ShuffleNetV2 while achieving smaller drops in accuracies compared to previous parameter-sharing and pruning approaches. We further show that the proposed approach is efficient in terms of latency and energy consumption on modern embedded devices.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.39 no.3
• /
• pp.141-148
• /
• 2021

This paper presents a method to detect zebra-crossing using deep learning which combines SegNet and ResNet. For the blind, a safe crossing system is important to know exactly where the zebra-crossings are. Zebra-crossing detection by deep learning can be a good solution to this problem and robotic vision-based assistive technologies sprung up over the past few years, which focused on specific scene objects using monocular detectors. These traditional methods have achieved significant results with relatively long processing times, and enhanced the zebra-crossing perception to a large extent. However, running all detectors jointly incurs a long latency and becomes computationally prohibitive on wearable embedded systems. In this paper, we propose a model for fast and stable segmentation of zebra-crossing from captured images. The model is improved based on a combination of SegNet and ResNet and consists of three steps. First, the input image is subsampled to extract image features and the convolutional neural network of ResNet is modified to make it the new encoder. Second, through the SegNet original up-sampling network, the abstract features are restored to the original image size. Finally, the method classifies all pixels and calculates the accuracy of each pixel. The experimental results prove the efficiency of the modified semantic segmentation algorithm with a relatively high computing speed.

• Journal of Broadcast Engineering
• /
• v.27 no.6
• /
• pp.840-847
• /
• 2022

In a deep neural network (DNN)-based speech enhancement, using global and local input speech information is closely related to model performance. Recently, a nested U-Net structure that utilizes global and local input data information using multi-scale has bee n proposed. This nested U-Net was also applied to speech enhancement and showed outstanding performance. However, a single skip connection used in nested U-Nets must be modified for the nested structure. In this paper, we propose a multi-level skip connection (MLS) to optimize the performance of the nested U-Net-based speech enhancement algorithm. As a result, the proposed MLS showed excellent performance improvement in various objective evaluation metrics compared to the standard skip connection, which means th at the MLS can optimize the performance of the nested U-Net-based speech enhancement algorithm. In addition, the final proposed m odel showed superior performance compared to other DNN-based speech enhancement models.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2016.05a
• /
• pp.97-100
• /
• 2016

Using the CNN(Convolutinal Neural Network), Deep Learning for variety of fields are being developed and these are showing significantly high level of performance at image recognition field. In this paper, we show the test accuracy which is learned by large-scale training data, over 5,000,000 of Korean characters. The architecture of CNN used in this paper is KCR(Korean Character Recognition)-AlexNet newly created based on AlexNet. KCR-AlexNet finally showed over 98% of test accuracy. The experimental data used in this paper is large-scale Korean character database PHD08 which has 2,187 samples for each Korean character and there are 2,350 Korean characters that makes total 5,139,450 sample data. Through this study, we show the excellence of architecture of KCR-AlexNet for learning PHD08.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.43 no.4
• /
• pp.511-523
• /
• 2023

Research on the integration of unmanned aerial vehicles and deep learning for reinforced concrete damage detection is actively underway. Convolutional neural networks have a high impact on the performance of image classification, detection, and segmentation as backbones. The MobileNet, a pre-trained convolutional neural network, is efficient as a backbone for an unmanned aerial vehicle-based damage detection model because it can achieve sufficient accuracy with low computational complexity. Analyzing vanilla convolutional neural networks and MobileNet under various conditions, MobileNet was evaluated to have a verification accuracy 6.0~9.0% higher than vanilla convolutional neural networks with 15.9~22.9% lower computational complexity. MobileNetV2, MobileNetV3Large and MobileNetV3Small showed almost identical maximum verification accuracy, and the optimal conditions for MobileNet's reinforced concrete damage image feature extraction were analyzed to be the optimizer RMSprop, no dropout, and average pooling. The maximum validation accuracy of 75.49% for 7 types of damage detection based on MobilenetV2 derived in this study can be improved by image accumulation and continuous learning.

• Nuclear Engineering and Technology
• /
• v.52 no.12
• /
• pp.2709-2716
• /
• 2020

Compositions of large nuclear cores (e.g. boiling water reactors) are highly heterogeneous in terms of fuel composition, control rod insertions and flow regimes. For this reason, they usually lack high order of symmetry (e.g. 1/4, 1/8) making it difficult to estimate their neutronic parameters for large spaces of possible loading patterns. A detailed hyperparameter optimization technique (a combination of manual and Gaussian process search) is used to train and optimize deep neural networks for the prediction of three neutronic parameters for the Ringhals-1 BWR unit: power peaking factors (PPF), control rod bank level, and cycle length. Simulation data is generated based on half-symmetry using PARCS core simulator by shuffling a total of 196 assemblies. The results demonstrate a promising performance by the deep networks as acceptable mean absolute error values are found for the global maximum PPF (~0.2) and for the radially and axially averaged PPF (~0.05). The mean difference between targets and predictions for the control rod level is about 5% insertion depth. Lastly, cycle length labels are predicted with 82% accuracy. The results also demonstrate that 10,000 samples are adequate to capture about 80% of the high-dimensional space, with minor improvements found for larger number of samples. The promising findings of this work prove the ability of deep neural networks to resolve high dimensionality issues of large cores in the nuclear area.

• Journal of the Korea Society of Computer and Information
• /
• v.25 no.5
• /
• pp.11-18
• /
• 2020

This study proposes a technique for the determination of outlet blockage using deep learning in a conveyor system. The proposed method aims to apply the best model to the actual process, where we train various CNN models for the determination of outlet blockage using images collected by CCTV in an industrial scene. We used the well-known CNN model such as VGGNet, ResNet, DenseNet and NASNet, and used 18,000 images collected by CCTV for model training and performance evaluation. As a experiment result with various models, VGGNet showed the best performance with 99.03% accuracy and 29.05ms processing time, and we confirmed that VGGNet is suitable for the determination of outlet blockage.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.4
• /
• pp.1246-1262
• /
• 2021

Recent studies have demonstrated the strong ability of deep convolutional neural networks (CNNs) to significantly boost the performance in single image super-resolution (SISR). The key concern is how to efficiently recover and utilize diverse information frequencies across multiple network layers, which is crucial to satisfying super-resolution image reconstructions. Hence, previous work made great efforts to potently incorporate hierarchical frequencies through various sophisticated architectures. Nevertheless, economical SISR also requires a capable structure design to balance between restoration accuracy and computational complexity, which is still a challenge for existing techniques. In this paper, we tackle this problem by proposing a competent architecture called Enhanced U-Net Network (EUN), which can yield ready-to-use features in miscellaneous frequencies and combine them comprehensively. In particular, the proposed building block for EUN is enhanced from U-Net, which can extract abundant information via multiple skip concatenations. The network configuration allows the pipeline to propagate information from lower layers to higher ones. Meanwhile, the block itself is committed to growing quite deep in layers, which empowers different types of information to spring from a single block. Furthermore, due to its strong advantage in distilling effective information, promising results are guaranteed with comparatively fewer filters. Comprehensive experiments manifest our model can achieve favorable performance over that of state-of-the-art methods, especially in terms of computational efficiency.