• 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 KIISE:Software and Applications
• /
• v.30 no.9
• /
• pp.829-842
• /
• 2003

In this Paper, we propose the new face detection and tracking method based on template matching for real-time applications such as, teleconference, telecommunication, front stage of surveillance system using face recognition, and video-phone applications. Since the main purpose of paper is to track a face regardless of various environments, we use template-based face tracking method. To generate robust face templates, we apply wavelet transform to the average face image and extract three types of wavelet template from transformed low-resolution average face. However template matching is generally sensitive to the change of illumination conditions, we apply Min-max normalization with histogram equalization according to the variation of intensity. Tracking method is also applied to reduce the computation time and predict precise face candidate region. Finally, facial components are also detected and from the relative distance of two eyes, we estimate the size of facial ellipse.

• Journal of the Korean Academy of Child and Adolescent Psychiatry
• /
• v.22 no.1
• /
• pp.25-37
• /
• 2011

Objectives：This study was conducted to develop a Korean algorithm of pharmacological and non-pharmacological treatment strategies in attention-deficit hyperactivity disorder (ADHD) and its specific comorbid disorders (e.g. tic disorder, depressive disorder, anxiety disorder, bipolar disorder, and oppositional defiant disorder/conduct disorder). Methods：Based on a literature review and expert consensus, both paper- and web-based survey tools were developed with respect to a comprehensive range of questions. Most options were scored using a 9-point scale for rating the appropriateness of medical decisions. For the other options, the surveyed experts were asked to provide answers (e.g., duration of treatment, aver-age dosage) or check boxes to indicate their preferred answers. The survey was performed on-line in a self-administered manner. Ultimately, 49 Korean child & adolescent psychiatrists, who had been considered experts in the treatment of ADHD, vol untarily completed the questionnaire. In analyzing the responses to items rated using the 9-point scale, consensus on each option was defined as a non-random distribution of scores as determined by a chi-square test. We assigned a categorical rank (first line/preferred choice, second line/alternate choice, third line/usually inappropriate) to each option based on the 95% confidence interval around the mean rating score. Results：Specific medication strategies for key clinical situations in ADHD and its comorbid disorders were indicated and described. We organized the suggested algorithms of ADHD treatment mainly on the basis of the opinions of the Korean experts. The suggested algorithm was constructed according to the templates of the Texas Child & Adolescent medication algorithm Project (CMAP). Conclusion：We have proposed a Korean treatment algorithm for ADHD, both with and without comorbid disorders through expert consensus and a broad literature review. As the tools available for ADHD treatment evolve, this algorithm could be reorganized and modified as required to suit updated scientific and clinical research findings.