• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.3
• /
• pp.89-100
• /
• 2023

Recently, on-device artificial intelligence (AI) solutions using mobile devices and embedded edge devices have emerged in various fields, such as computer vision, to address network traffic burdens, low-energy operations, and security problems. Although vision transformer deep learning models have outperformed conventional convolutional neural network (CNN) models in computer vision, they require more computations and parameters than CNN models. Thus, they are not directly applicable to embedded edge devices with limited hardware resources. Many researchers have proposed various model compression methods or lightweight architectures for vision transformers; however, there are only a few studies evaluating the effects of model compression techniques of vision transformers on performance. Regarding this problem, this paper presents a performance evaluation of vision transformers on embedded platforms. We investigated the behaviors of three vision transformers: DeiT, LeViT, and MobileViT. Each model performance was evaluated by accuracy and inference time on edge devices using the ImageNet dataset. We assessed the effects of the quantization method applied to the models on latency enhancement and accuracy degradation by profiling the proportion of response time occupied by major operations. In addition, we evaluated the performance of each model on GPU and EdgeTPU-based edge devices. In our experimental results, LeViT showed the best performance in CPU-based edge devices, and DeiT-small showed the highest performance improvement in GPU-based edge devices. In addition, only MobileViT models showed performance improvement on EdgeTPU. Summarizing the analysis results through profiling, the degree of performance improvement of each vision transformer model was highly dependent on the proportion of parts that could be optimized in the target edge device. In summary, to apply vision transformers to on-device AI solutions, either proper operation composition and optimizations specific to target edge devices must be considered.

• International Journal of Computer Science & Network Security
• /
• v.23 no.4
• /
• pp.95-102
• /
• 2023

The emergence of Internet of Things (IoT) into our daily lives has grown rapidly. It's been integrated to our homes, cars, and cities, increasing the intelligence of devices involved in communications. Enormous amount of data is exchanged over smart devices through the internet, which raises security concerns in regards of privacy evasion. This paper is focused on the forensics and intrusion detection on one of the most common protocols in IoT environments, especially smart home environments, which is the Message Queuing Telemetry Transport (MQTT) protocol. The paper covers general IoT infrastructure, MQTT protocol and attacks conducted on it, and multiple network forensics frameworks in smart homes. Furthermore, a machine learning model is developed and tested to detect several types of attacks in an IoT network. A forensics tool (MQTTracker) is proposed to contribute to the investigation of MQTT protocol in order to provide a safer technological future in the warmth of people's homes. The MQTT-IOT-IDS2020 dataset is used to train the machine learning model. In addition, different attack detection algorithms are compared to ensure the suitable algorithm is chosen to perform accurate classification of attacks within MQTT traffic.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.46 no.2
• /
• pp.72-81
• /
• 2023

This research proposes a novel approach to tackle the challenge of categorizing unstructured customer complaints in the automotive industry. The goal is to identify potential vehicle defects based on the findings of our algorithm, which can assist automakers in mitigating significant losses and reputational damage caused by mass claims. To achieve this goal, our model uses the Word2Vec method to analyze large volumes of unstructured customer complaint data from the National Highway Traffic Safety Administration (NHTSA). By developing a score dictionary for eight pre-selected criteria, our algorithm can efficiently categorize complaints and detect potential vehicle defects. By calculating the score of each complaint, our algorithm can identify patterns and correlations that can indicate potential defects in the vehicle. One of the key benefits of this approach is its ability to handle a large volume of unstructured data, which can be challenging for traditional methods. By using machine learning techniques, we can extract meaningful insights from customer complaints, which can help automakers prioritize and address potential defects before they become widespread issues. In conclusion, this research provides a promising approach to categorize unstructured customer complaints in the automotive industry and identify potential vehicle defects. By leveraging the power of machine learning, we can help automakers improve the quality of their products and enhance customer satisfaction. Further studies can build upon this approach to explore other potential applications and expand its scope to other industries.

• Journal of the Korean Society of Safety
• /
• v.38 no.6
• /
• pp.50-59
• /
• 2023

As the diversity of disasters continues to increase, the concept of "complex disasters" has gained prominence in various policies and studies related to disaster management. However, there has been a certain limitation in the availability of the systematic statistics or data in advancing policies and research initiatives related to complex disasters. This study aims to analyze the macro-level characteristics of the complex disasters that have occurred domestically utilizing a 30-year span of a news data. Initially, we categorize the complex disasters into the three types: "Natural disaster-Natural disaster", "Natural disaster-Social disaster", and "Social disaster-Social disaster". As a result, the "natural diaster-social disaster" type is the most prevalent. It is noted that "natual disaster-natural disaster" type has increased significantly in recent 10 years (2011-2020). In terms of specific disaster types, "Storm and Flood", "Collapse", "Traffic Accident", "National Infrastructure Paralysis", and "Fire⋅Explosion" occur the most in conjunction with other disasters in a complex manner. It has been observed that the types of disasters co-ocuuring with others have become more diverse over time. Parcicularly, in recent 10 years (2011-2020), in addition to the aforementioned five types, "Heat Wave", "Heavy Snowfall⋅Cold Wave", "Earthquake", "Chemical Accident", "Infectious Disease", "Forest Fire", "Air Pollution", "Drought", and "Landslide" have been notable for their frequent co-occurrence with other disasters. These findings through the statistical analysis of the complex disasters using long-term news data are expected to serve as crucial data for future policy development and research on complex disaster management.

• Archives of Plastic Surgery
• /
• v.50 no.6
• /
• pp.550-556
• /
• 2023

Blunt breast trauma occurs in 2% of blunt chest injuries. This study aimed to evaluate the evidence on breast reconstruction after blunt trauma associated with the use of a seat belt. Also, we describe the first case of breast reconstruction using the Ribeiro technique. In November 2022, a systematic search of MEDLINE, EMBASE, and Google Scholar databases was conducted. The literature was screened independently by two reviewers, and the data was extracted. Our search terms included breast, mammoplasty, blunt injury, and seat belts. In addition, we present the case of a woman with a left breast deformity and her reconstruction using the inferior Ribeiro flap technique. Six articles were included. All included studies were published between 2010 and 2021. The studies recruited seven patients. According to the Teo and Song classification, seven class 2b cases were reported. In five cases a breast reduction was performed in the deformed breast with different types of pedicles (three superomedial flaps, one lower flap, one superior flap). Only one case presented complications. The case here presented was a type 2b breast deformity in which the lower Ribeiro pedicle was used successfully without complications during follow-up. Until now there has been no consensus on reconstructive treatment due to the rarity of this entity. However, we must consider surgical treatment individually for each patient. We believe that the Ribeiro technique is a feasible and safe alternative in the treatment of posttraumatic breast deformities, offering very good long-term results.

• Journal of Korean Orthopaedic Sports Medicine
• /
• v.11 no.1
• /
• pp.44-50
• /
• 2012

Purpose: The purpose of this study is to analyze the relationship between acute anterior cruciate ligament (ACL) injury and bone bruise using the survey for location and incidence of bone bruise. Materials and Methods: From Jan. 2006 to Feb. 2010, 87 knees from who had complaint a traumatic knee pain were diagnosed as acute ACL tear using MRI evaluation. Associated injury, location and incidence of bone bruise were analyzed using MRI. The location of bone bruise on the MRI was classified as medial, central and lateral area on anteroposterior and lateral view of femur and tibia. The bone bruise was classified with Costa Paz classification. Results: Bone bruise of injury during daily living activity were located at medial area on coronary view and anterior area on sagittal view of distal femur, at medial area on coronary view and anterior area on sagittal view of proximal tibia (p=0.024, p=0.021, p=0.025 and p=0.029, respectively). Bone bruise of injury during sports activity were located at lateral area on coronary view and central area on sagittal view of distal femur, at lateral area on coronary view and posterior area on sagittal view of proximal tibia (p=0.014, p=0.015, p=0.018 and p=0.017, respectively). Bone bruise patterns due to traffic accident were inconclusive (p=0.264, p=0.254, p=0.229 and p=0.267, respectively). Conclusion: Injury mechanism of acute ACL injury from activities of daily living or sports activities compared to that of traffic accident showed a more consistent bone bruise patterns. Special attention to acute ACL tear must be paid in case of bone bruise at lateral tibial plateau and lateral femoral condyle.

• International Journal of Highway Engineering
• /
• v.11 no.4
• /
• pp.69-78
• /
• 2009

Korea has achieved significant economic growth with building the Gyeongbu Expressway. As the number of new road construction projects has decreased, it becomes more important to maintain optimal status of the current road networks. One of the best ways to accomplish it is weight enforcement as active control measure of traffic load. This study is to develop High-speed Weigh-in-motion System in order to enhance efficiency of weight enforcement, and to analyze patterns of overloaded trucks on highways through the system. Furthermore, it is to review possibilities of developing overweight control system with application of the HS-WIM system. The HS-WIM system developed by this study consists of two sets of an axle load sensor, a loop sensor and a wandering sensor on each lane. A wandering sensor detects whether a travelling vehicle is off the lane or not with the function of checking the location of tire imprint. The sensor of the WIM system has better function of classifying types of vehicles than other existing systems by detecting wheel distance and tire type such as single or dual tire. As a result, its measurement errors regarding 12 types of vehicle classification are very low, which is an advantage of the sensor. The verification tests of the system under all conditions showed that the mean measurement errors of axle weight and gross axle weight were within 15 percent and 7 percent respectively. According to the WIM rate standard of the COST-323, the WIM system of this study is ranked at B(10). It means the system is appropriate for the purpose of design, maintenance and valuation of road infrastructure. The WIM system in testing a 5-axle cargo truck, the most frequently overloaded vehicle among 12 types of vehicles, is ranked at A(5) which means the system is available to control overloaded vehicles. In this case, the measurement errors of axle load and gross axle load were within 8 percent and 5 percent respectively. Weight analysis of all types of vehicles on highways showed that the most frequently overloaded vehicles were type 5, 6, 7 and 12 among 12 vehicle types. As a result, it is necessary to use more effective overweight enforcement system for vehicles which are seriously overloaded due to their lift axles. Traffic volume data depending upon vehicle types is basic information for road design and construction, maintenance, analysis of traffic flow, road policies as well as research.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.26 no.2
• /
• pp.139-146
• /
• 2013

This study is to update of vibration power generating model and to analyze electromotive force on subway. Analysis of electromotive force using power generation depending on classification of locations which are ballast bed and concrete bed. As the section between Seocho and Bangbae in the line 2 subway was changed from ballast bed to concrete bed, it could be analyzed at same condition, train, section. Induced electromotive force equation by Faraday's law was updated using Bayesian regression and correlation analysis with calculate value and experiment value. Using the updated model, it could get 40mV per one power generation in ballast bed, and it also could get 4mV per one power generation in concrete bed. If the updated model apply to subway or any train, it will be more effective to get electric power. In addition to that, it will be good to reduce greenhouse gas and to build a green traffic network.

• Journal of Korean Foot and Ankle Society
• /
• v.16 no.2
• /
• pp.94-100
• /
• 2012

Purpose: The purpose of this study is to analyze the clinical features of distal tibia fractures and to evaluate the treatment outcomes of minimally invasive plate osteosynthesis (MIPO). Materials and Methods: From January 2004 to December 2009, 84 cases of 81 patients treated with plate fixation for distal tibia fracture were enrolled in this retrospective review. We investigated age, sex, injury mechanism, fracture patterns, and complications, and the clinical features were analyzed. To evaluate the treatment outcomes of MIPO, we divided into two groups. MIPO group consisted of 55 patients were treated with MIPO technique and conventional group consisted of 18 patients were treated with open reduction and internal fixation with conventional anterolateral plating. The results were compared between two groups by assessing bony union time, operation time, amount of blood loss, range of ankle motion, clinical score by American Orthopaedic Foot and Ankle Society (AOFAS) score, and post-operative complications. Results: The mean age of 81 patients with distal tibia fracture was 54.8 years. According to AO classification, A1:2:3 were 16, 20, 16 patients, B1:2:3 were 2, 8, 7, C1:2:3 were 1, 3, 11 patients. According to injury mechanism, slip down injury was patients, traffic accident was 26, fall from height injury was 14 patients respectively. The type A fractures were lower energy trauma and more older patients. The type C fractures were higher energy trauma and younger patients. MIPO group was better than conventional group in operative time, blood loss, bony union time, and ankle joint motion. In complications, MIPO group showed no nonunion and infection, one malunion, one skin necrosis, nine skin irritations, and one screw breakage. Conventional group showed two nonunion, four infections, two skin necrosis, and one metal failure. Conclusion: Distal tibial fractures caused by low energy trauma were on the increase. Minimal invasive plate osteosynthesis was shorter bony union time and operation time, less blood loss, and larger ankle motions than conventional open reduction and plate fixation.

• Journal of Trauma and Injury
• /
• v.25 no.1
• /
• pp.1-6
• /
• 2012

Purpose: This study analyzed the characteristics of unstable pelvic bone fractures associated with intra-abdominal solid organ injury. Methods: Medical records were retrospectively collected from January 2000 to December 2010 for patients with unstable pelvic bone fractures. Unstable pelvic bone fracture was defined as lateral compression types II and III, antero-posterior compression types II and III, vertical shear and combined type by young classification. Subjects were divided into two groups, with (injured group) and without (non-injured group) intra-abdominal solid organ injury, to evaluate whether the characteristics of the fractured depended on the presence of associated solid organ injury. Data included demographics, mechanism of injury, initial hemodynamic status, laboratory results, revised trauma score (RTS), abbreviated injury scale (AIS), injury severity score (ISS), amount of transfusion, admission to the intensive care unit (ICU), and mortality. Results: The subjects were 217 patients with a mean age of 44 years and included 134 male patients(61.8%). The injured group included 38 patients(16.9%). Traffic accidents were the most common mechanism of injury, and lateral compression was the most common type of fracture in all groups. The initial blood pressure was lower in the injured group, and the ISS was greater. The arterial pH was lower in the injured group, and shock within 24 hours after arrival at the emergency department was more frequent in the injured group. The amount of the transfused packed red blood cells within 24 hours was higher in the injured group than the non-injured group. Invasive treatment, including surgery and angiographic embolization, was more common in the injured group, and the stay in the ICU was longer in the injured group. Conclusion: A need exists to decide on a diagnostic and therapeutic plan regarding the possibility of intra-abdominal solid organ injury for hemodynamically unstable patients with unstable pelvic bone fractures and multiple associated injuries.