• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.6
• /
• pp.319-325
• /
• 2023

In this study, we proposed a method for hyperparameter optimization in the building and training of a deep learning model designed to process point cloud data collected by a millimeter-wave radar system. The primary aim of this study is to facilitate the deployment of a baseline model in resource-constrained IoT devices. We evaluated a RadHAR baseline deep learning model trained on a public dataset composed of point clouds representing five distinct human activities. Additionally, we introduced a coarse-to-fine hyperparameter optimization procedure, showing substantial potential to enhance model efficiency without compromising predictive performance. Experimental results show the feasibility of significantly reducing model size without adversely impacting performance. Specifically, the optimized model demonstrated a 3.3% improvement in classification accuracy despite a 16.8% reduction in number of parameters compared th the baseline model. In conclusion, this research offers valuable insights for the development of deep learning models for resource-constrained IoT devices, underscoring the potential of hyperparameter optimization and model size reduction strategies. This work contributes to enhancing the practicality and usability of deep learning models in real-world environments, where high levels of accuracy and efficiency in data processing and classification tasks are required.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.6
• /
• pp.303-310
• /
• 2023

A single vision system limits the ability to accurately understand the spatial constraints and interactions between robots and dynamic workers caused by gantry robots and collaborative robots during production manufacturing. In this paper, we propose a 3D pose registration method for dynamic workers based on a multi-domain vision system for safety monitoring in manufacturing environments. This method uses OpenPose, a deep learning-based posture estimation model, to estimate the worker's dynamic two-dimensional posture in real-time and reconstruct it into three-dimensional coordinates. The 3D coordinates of the reconstructed multi-domain vision system were aligned using the ICP algorithm and then registered to a single 3D coordinate system. The proposed method showed effective performance in a manufacturing process environment with an average registration error of 0.0664 m and an average frame rate of 14.597 per second.

• Asia pacific journal of information systems
• /
• v.29 no.1
• /
• pp.35-49
• /
• 2019

The paper attempts to document the application of relevant Machine Learning (ML) models on Yelp (a crowd-sourced local business review and social networking site) dataset to analyze, predict and recommend business. Strategically using two cloud platforms to minimize the effort and time required for this project. Seven machine learning algorithms in Azure ML of which four algorithms are implemented in Databricks Spark ML. The analyzed Yelp business dataset contained 70 business attributes for more than 350,000 registered business. Additionally, review tips and likes from 500,000 users have been processed for the project. A Recommendation Model is built to provide Yelp users with recommendations for business categories based on their previous business ratings, as well as the business ratings of other users. Classification Model is implemented to predict the popularity of the business as defining the popular business to have stars greater than 3 and unpopular business to have stars less than 3. Text Analysis model is developed by comparing two algorithms, uni-gram feature extraction and n-feature extraction in Azure ML studio and logistic regression model in Spark. Comparative conclusions have been made related to efficiency of Spark ML and Azure ML for these models.

• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.774-776
• /
• 2024

The effective implementation of advanced speech recognition (ASR) systems necessitates the deployment of sophisticated keyword spotting models that are both responsive and resource-efficient. The initial local detection of user interactions is crucial as it allows for the selective transmission of audio data to cloud services, thereby reducing operational costs and mitigating privacy risks associated with continuous data streaming. In this paper, we address these needs and propose utilizing the Whisper-Tiny model with fine-tuning process to specifically recognize keywords from google speech dataset which includes 65000 audio clips of keyword commands. By adapting the model's encoder and appending a lightweight classification head, we ensure that it operates within the limited resource constraints of local devices. The proposed model achieves the notable test accuracy of 92.94%. This architecture demonstrates the efficiency as on-device model with stringent resources leading to enhanced accessibility in everyday speech recognition applications.

• Journal of Platform Technology
• /
• v.12 no.1
• /
• pp.77-84
• /
• 2024

With the recent increase in the use of artificial intelligence, AI TRiSM data management within organizations has become important, and thus securing data reliability has emerged as an essential requirement for data-based decision-making. Digital content is transmitted through the unreliable Internet to the cloud where the digital content storage is located, then used in various applications. When detecting anomaly of data, it is difficult to provide a function to check content modification due to its damage in digital content systems. In this paper, we design a technique to guarantee the reliability of video data by expanding the function of data annotation. The designed annotation technique constitutes a prototype based on gRPC to handle a request and a response in a webUI that generates classification label and Merkle tree of given video data.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.8
• /
• pp.2381-2398
• /
• 2024

The construction of virtual indoor spaces is crucial for the development of metaverses, virtual production, and other 3D content domains. Traditional methods for creating these spaces are often cost-prohibitive and labor-intensive. To address these challenges, we present a pipeline for generating digital twins of real indoor environments from RGB-D camera-scanned data. Our pipeline synergizes space structure estimation, 3D object detection, and the inpainting of missing areas, utilizing deep learning technologies to automate the creation process. Specifically, we apply deep learning models for object recognition and area inpainting, significantly enhancing the accuracy and efficiency of virtual space construction. Our approach minimizes manual labor and reduces costs, paving the way for the creation of metaverse spaces that closely mimic real-world environments. Experimental results demonstrate the effectiveness of our deep learning applications in overcoming traditional obstacles in digital twin creation, offering high-fidelity digital replicas of indoor spaces. This advancement opens for immersive and realistic virtual content creation, showcasing the potential of deep learning in the field of virtual space construction.

• The KIPS Transactions:PartD
• /
• v.19D no.2
• /
• pp.187-202
• /
• 2012

As a representative reuse paradigm, the theme of service-oriented Paradigm (SOC) is largely centered on publishing and subscribing reusable services. Here, SOC is the term including service oriented architecture and cloud computing. Service providers can produce high profits with reusable services, and service consumers can develop their applications with less time and effort by reusing the services. Design Patterns (DP) is a set of reusable methods to resolve commonly occurring design problems and to provide design structures to deal with the problems by following open/close princples. However, since DPs are mainly proposed for building object-oriented systems and there are distinguishable differences between object-oriented paradigm and SOC, it is challenging to apply the DPs to SOC design problems. Hence, DPs need to be customized by considering the two aspects; for service providers to design services which are highly reusable and reflect their unique characteristics and for service consumers to develop their target applications by reusing and customizing services as soon as possible. Therefore, we propose a set of DPs that are customized to SOC. With the proposed DPs, we believe that service provider can effectively develop highly reusable services, and service consumers can efficiently adapt services for their applications.

• Journal of the Korean Data and Information Science Society
• /
• v.26 no.5
• /
• pp.1129-1139
• /
• 2015

Inference process generates additional triples from knowledge represented in RDF triples of semantic web technology. Tens of million of triples as an initial big data and the additionally inferred triples become a knowledge base for applications such as QA(question&answer) system. The inference engine requires more computing resources to process the triples generated while inferencing. The additional computing resources supplied by underlying resource pool in cloud computing can shorten the execution time. This paper addresses an algorithm to allocate the number of computing nodes "elastically" at runtime on Hadoop, depending on the size of knowledge data fed. The model proposed in this paper is composed of the layered architecture: the top layer for applications, the middle layer for distributed parallel inference engine to process the triples, and lower layer for elastic Hadoop and server visualization. System algorithms and test data are analyzed and discussed in this paper. The model hast the benefit that rich legacy Hadoop applications can be run faster on this system without any modification.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.3
• /
• pp.91-101
• /
• 2024

The utilization of container virtualization technology ensures the consistency and portability of data-intensive and memory volatile workflows. Kubernetes serves as the de facto standard for orchestrating these container applications. Cloud users often overprovision container applications to avoid container restarts caused by resource shortages. However, overprovisioning results in decreased CPU and memory resource utilization. To address this issue, oversubscription of container resources is commonly employed, although excessive oversubscription of memory resources can lead to a cascade of container restarts due to node memory scarcity. Container restarts can reset operations and impose substantial overhead on containers with high memory volatility that include numerous stateful applications. This paper proposes a technique to mitigate container restarts in a memory oversubscription environment based on Kubernetes. The proposed technique involves identifying containers that are likely to request memory allocation on nodes experiencing high memory usage and temporarily pausing these containers. By significantly reducing the CPU usage of containers, an effect similar to a paused state is achieved. The suspension of the identified containers is released once it is determined that the corresponding node's memory usage has been reduced. The average number of container restarts was reduced by an average of 40% and a maximum of 58% when executing a high memory volatile workflow in a Kubernetes environment with the proposed method compared to its absence. Furthermore, the total execution time of a container workflow is decreased by an average of 7% and a maximum of 13% due to the reduced frequency of container restarts.

• Journal of Astronomy and Space Sciences
• /
• v.26 no.4
• /
• pp.555-566
• /
• 2009

Precipitable water vapor (PWV) derived from a numerical weather prediction (NWP) model were compared to observations derived from ground-based Global Positioning System (GPS) receivers. The model data compared were from the Weather Research and Forecasting (WRF) model short-range forecasts on nested grids. The numerical experimets were performed by selecting the cloud microphysics schemes and for the comparisons, the Changma period of 2008 was selected. The observational data were derived from GPS measurements at 9-sites in South Korea over a 1-month period, in the middle of June-July 2008. In general, the WRF model demonstrated considerable skill in reproducing the temporal and spatial evolution of the PWV as depicted by the GPS estimations. The correlation between forecasts and GPS estimates of PWV depreciated slowly with increasing forecast times. Comparing simulations with a resolution of 18 km and 6 km showed no obvious PWV dependence on resolution. Besides, GPS and the model PWV data were found to be in quite good agreement with data derived from radiosondes. These results indicated that the GPS-derived PWV data, with high temporal and spatial resolution, are very useful for meteorological applications.