Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049937
Hosung Son, Suk-ju Kang
Depth estimation methods based on deep learning have been studied to improve depth estimation accuracy. However, obtaining inter-frame consistency in depth maps in video depth estimation remains a challenge. Therefore, we proposed an application methodology for spatio-temporal consistency enhancement in video depth estimation based on convolutional neural networks (CNNs) and recurrent neural networks (RNNs). In other words, the convolutional long-short term memory (ConvLSTM) module was added to the decoder of depth estimation network to enable the use of the information from the previous frames. Additionally, the one-stage learning process was implemented to ensure ease of training. In conclusion, we experimentally show that the proposed method can achieve not only improved accuracy also consistency between depth map frames.
{"title":"Multi-view stereo with recurrent neural networks for spatio-temporal consistent depth maps","authors":"Hosung Son, Suk-ju Kang","doi":"10.1109/ICEIC57457.2023.10049937","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049937","url":null,"abstract":"Depth estimation methods based on deep learning have been studied to improve depth estimation accuracy. However, obtaining inter-frame consistency in depth maps in video depth estimation remains a challenge. Therefore, we proposed an application methodology for spatio-temporal consistency enhancement in video depth estimation based on convolutional neural networks (CNNs) and recurrent neural networks (RNNs). In other words, the convolutional long-short term memory (ConvLSTM) module was added to the decoder of depth estimation network to enable the use of the information from the previous frames. Additionally, the one-stage learning process was implemented to ensure ease of training. In conclusion, we experimentally show that the proposed method can achieve not only improved accuracy also consistency between depth map frames.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132735691","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049967
Eung-Ku Kim, Nahyun Lee, Yoon-Sik Cho
Large-scale vision-language representation learning has shown dramatic advances in image-text retrieval studies. On top of that, it is clear that the CLIP (Contrastive Language-Image Pre-training) [1] has contributed to this field in that it succeeded pre-training on tremendous web multimedia data with leveraging contrastive loss, one of the most competitive loss for multimodal learning. CLIP4CLIP [2] utilized the pre-trained CLIP weight to appeal its performance of learning video representations with textual supervision with additional post-pretraining on Howto100M [3]. However, there is still room for elevation in CLIP4Clip from perspective of contrastive learning - large enough instances of negative samples help to learn better representations, as the batch size in each training batch getting larger, the more chance of the model to be allowed to contrast more semantic representations. In the same queue of this point of view, we propose CwMTVR combining Momentum Contrast (MoCo) [4], cross-modal momentum contrastive learning framework into CLIP4Clip. Experimental results prove that the CwMTVR model achieve competitive results in text-video retrieval on MSVD dataset.
{"title":"A Unified framework based on Large-scale Momentum Contrastive learning for Text-Video Retrieval","authors":"Eung-Ku Kim, Nahyun Lee, Yoon-Sik Cho","doi":"10.1109/ICEIC57457.2023.10049967","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049967","url":null,"abstract":"Large-scale vision-language representation learning has shown dramatic advances in image-text retrieval studies. On top of that, it is clear that the CLIP (Contrastive Language-Image Pre-training) [1] has contributed to this field in that it succeeded pre-training on tremendous web multimedia data with leveraging contrastive loss, one of the most competitive loss for multimodal learning. CLIP4CLIP [2] utilized the pre-trained CLIP weight to appeal its performance of learning video representations with textual supervision with additional post-pretraining on Howto100M [3]. However, there is still room for elevation in CLIP4Clip from perspective of contrastive learning - large enough instances of negative samples help to learn better representations, as the batch size in each training batch getting larger, the more chance of the model to be allowed to contrast more semantic representations. In the same queue of this point of view, we propose CwMTVR combining Momentum Contrast (MoCo) [4], cross-modal momentum contrastive learning framework into CLIP4Clip. Experimental results prove that the CwMTVR model achieve competitive results in text-video retrieval on MSVD dataset.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125802667","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049848
Dajin Han, Tae Hyun Kim
Normalizing Flow (NF) has gained growing popularity in various image generation tasks. In this work, we develop a new method that enables the NF to control face generation, which has not been studied yet. To do so, we introduce several loss functions to facilitate stable training and inference while controlling face generation given a facial landmark. In our experiments, we evaluate the performance of the proposed method and show the capability of the NF in controlling the face generation task.
{"title":"One-Shot Face Reenactment with 2D Facial Landmark Conditional Normalizing Flow","authors":"Dajin Han, Tae Hyun Kim","doi":"10.1109/ICEIC57457.2023.10049848","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049848","url":null,"abstract":"Normalizing Flow (NF) has gained growing popularity in various image generation tasks. In this work, we develop a new method that enables the NF to control face generation, which has not been studied yet. To do so, we introduce several loss functions to facilitate stable training and inference while controlling face generation given a facial landmark. In our experiments, we evaluate the performance of the proposed method and show the capability of the NF in controlling the face generation task.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122175015","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049884
Junho Kang, Soungchul Park, Suhwan Kim, Jooyeol Rhee
In this paper, we propose a programmable gain amplifier (PGA) for an ultrasonic sensor system that removes DC offset and out-of-band noise through band-pass filtering in automotive applications. The proposed PGA consists of a low-noise amplifier (LNA), a variable gain amplifier (VGA), an RC low-pass filter (LPF), and a voltage buffer. The structure of LNA and VGA is an active band-pass filter (BPF) constructed using an operational transconductance amplifier (OTA) and passive elements. The PGA’s voltage gain and cut-off frequency can be adjusted by controlling the resistor connected to the common-mode voltage, the feedback capacitor, and the OTA compensation capacitor. An AC coupling capacitor is placed at the input stage to eliminate the DC offset and low-frequency noise of the PGA input signal. In addition, by arranging the feedback resistor and the feedback capacitor, the band-pass filtering function is implemented to block out noise at frequencies higher than the signal band. The PGA can adjust the voltage gain with a 6-dB step from 56 dB to 110 dB in the ultrasonic signal band from 30 kHz to 70 kHz. The design is designed in a 0.18-µm CMOS process. The post-layout simulation shows that it consumes a current of 9.7 mA with a 5 V supply voltage.
{"title":"A 56-to-110 dB Gain Programmable Gain Amplifier with Second-Order Band Pass Filter for Ultrasonic Sensor Systems","authors":"Junho Kang, Soungchul Park, Suhwan Kim, Jooyeol Rhee","doi":"10.1109/ICEIC57457.2023.10049884","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049884","url":null,"abstract":"In this paper, we propose a programmable gain amplifier (PGA) for an ultrasonic sensor system that removes DC offset and out-of-band noise through band-pass filtering in automotive applications. The proposed PGA consists of a low-noise amplifier (LNA), a variable gain amplifier (VGA), an RC low-pass filter (LPF), and a voltage buffer. The structure of LNA and VGA is an active band-pass filter (BPF) constructed using an operational transconductance amplifier (OTA) and passive elements. The PGA’s voltage gain and cut-off frequency can be adjusted by controlling the resistor connected to the common-mode voltage, the feedback capacitor, and the OTA compensation capacitor. An AC coupling capacitor is placed at the input stage to eliminate the DC offset and low-frequency noise of the PGA input signal. In addition, by arranging the feedback resistor and the feedback capacitor, the band-pass filtering function is implemented to block out noise at frequencies higher than the signal band. The PGA can adjust the voltage gain with a 6-dB step from 56 dB to 110 dB in the ultrasonic signal band from 30 kHz to 70 kHz. The design is designed in a 0.18-µm CMOS process. The post-layout simulation shows that it consumes a current of 9.7 mA with a 5 V supply voltage.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125105393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049906
Yong Jin, M. Tomoishi, N. Yamai
Email is one of the important and indispensable Internet services, but in the meanwhile, the spread of emailderived phishing URLs has been one of the critical cyber threats for a long time. The security facilities in organization networks are suffering from monitoring and analyzing all traffic besides emails which consumes much computing resource. In this paper, we propose a trigger-based blocking mechanism for accessing the email-derived phishing URLs with user alert to protect the end users from phishing attacks. The proposed system practically uses the Domain Name System (DNS) and the Response Policy Zone (RPZ) feature to direct the triggered HTTP(S) access for the email-derived phishing URLs to a particular proxy. Then the HTTP(S) communication can be detected and blocked based on the users’ decisions by alerting them. A prototype for the proposed mechanism has been implemented and the preliminary feature evaluations in a local experimental network have been conducted. The evaluation results confirmed that all the HTTP(S) access for the email-derived phishing URLs was successfully directed to the pre-constructed particular HTTP(S) proxy, then an alert page was showed up to the end users, and the access was passed through or blocked based on the end users’ decisions.
{"title":"Trigger-based Blocking Mechanism for Access to Email-derived Phishing URLs with User Alert","authors":"Yong Jin, M. Tomoishi, N. Yamai","doi":"10.1109/ICEIC57457.2023.10049906","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049906","url":null,"abstract":"Email is one of the important and indispensable Internet services, but in the meanwhile, the spread of emailderived phishing URLs has been one of the critical cyber threats for a long time. The security facilities in organization networks are suffering from monitoring and analyzing all traffic besides emails which consumes much computing resource. In this paper, we propose a trigger-based blocking mechanism for accessing the email-derived phishing URLs with user alert to protect the end users from phishing attacks. The proposed system practically uses the Domain Name System (DNS) and the Response Policy Zone (RPZ) feature to direct the triggered HTTP(S) access for the email-derived phishing URLs to a particular proxy. Then the HTTP(S) communication can be detected and blocked based on the users’ decisions by alerting them. A prototype for the proposed mechanism has been implemented and the preliminary feature evaluations in a local experimental network have been conducted. The evaluation results confirmed that all the HTTP(S) access for the email-derived phishing URLs was successfully directed to the pre-constructed particular HTTP(S) proxy, then an alert page was showed up to the end users, and the access was passed through or blocked based on the end users’ decisions.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127407729","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049886
Khikmatullo Tulkinbekov, Deok-Hwan Kim
Blockchain protocol has emerged in modern technologies in recent years. Significantly, the data collection frameworks are one of the most demanding fields of blockchain adaptation. However, the traditional approach relies on heavy computations, which require high-computing servers that cannot be affordable by all providers. This paper proposes a reliable and secure data collection technique, BSave, that employs a lightweight blockchain suitable for low-computing devices. BSave stores the data in a key-value database and enables two-way encrypted security on both keys and values. Compared to traditional approaches, the BSave proved its effectiveness in data-intensive environments.
{"title":"Data Collection Framework Using a Lightweight Blockchain","authors":"Khikmatullo Tulkinbekov, Deok-Hwan Kim","doi":"10.1109/ICEIC57457.2023.10049886","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049886","url":null,"abstract":"Blockchain protocol has emerged in modern technologies in recent years. Significantly, the data collection frameworks are one of the most demanding fields of blockchain adaptation. However, the traditional approach relies on heavy computations, which require high-computing servers that cannot be affordable by all providers. This paper proposes a reliable and secure data collection technique, BSave, that employs a lightweight blockchain suitable for low-computing devices. BSave stores the data in a key-value database and enables two-way encrypted security on both keys and values. Compared to traditional approaches, the BSave proved its effectiveness in data-intensive environments.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129114558","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049878
Hoon-Seok Jang
Reconstructing a 3D shape using one or several images is one of the important factors in implementing a digital twin in the field of smart farm. Shape from Focus (SFF) is a passive method that reconstructs a 3D shape using 2D images having different focus levels. When 2D images are acquired at each step along the optical axis, mechanical vibrations occur. SFF techniques are vulnerable to jitter noise that changes the focus values of 2D images. In this manuscript, a new filtering technique that provides high accuracy and low computational cost for 3D shape recovery is proposed. First, jitter noise is modeled as a Lévy distribution. This assumption makes it possible to show the effectiveness of the proposed filtering technique in the presence of non-Gaussian noise. Second, the focus curves are modeled with a Gaussian function to compare the performance of the proposed filtering technique and the existing filtering techniques. Finally, the maximum correntropy criterion Kalman filter is designed and applied to the modeled focus curves. The experimental results demonstrate the effectiveness of proposed method.
利用一幅或多幅图像重建三维形状是实现智能农场领域数字孪生的重要因素之一。SFF (Shape from Focus)是一种利用不同聚焦水平的二维图像重建三维形状的被动方法。当沿着光轴的每一步获得二维图像时，就会发生机械振动。SFF技术容易受到改变二维图像焦点值的抖动噪声的影响。本文提出了一种高精度、低计算成本的三维形状恢复滤波技术。首先，将抖动噪声建模为lsamvy分布。这个假设使得在存在非高斯噪声的情况下显示所提出的滤波技术的有效性成为可能。其次，用高斯函数对焦点曲线进行建模，比较所提滤波技术与现有滤波技术的性能。最后，设计了最大熵准则卡尔曼滤波器，并将其应用于建模的聚焦曲线。实验结果证明了该方法的有效性。
{"title":"Acquisition of High-Quality Image by Using Maximum Correntropy Criterion Kalman Filter","authors":"Hoon-Seok Jang","doi":"10.1109/ICEIC57457.2023.10049878","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049878","url":null,"abstract":"Reconstructing a 3D shape using one or several images is one of the important factors in implementing a digital twin in the field of smart farm. Shape from Focus (SFF) is a passive method that reconstructs a 3D shape using 2D images having different focus levels. When 2D images are acquired at each step along the optical axis, mechanical vibrations occur. SFF techniques are vulnerable to jitter noise that changes the focus values of 2D images. In this manuscript, a new filtering technique that provides high accuracy and low computational cost for 3D shape recovery is proposed. First, jitter noise is modeled as a Lévy distribution. This assumption makes it possible to show the effectiveness of the proposed filtering technique in the presence of non-Gaussian noise. Second, the focus curves are modeled with a Gaussian function to compare the performance of the proposed filtering technique and the existing filtering techniques. Finally, the maximum correntropy criterion Kalman filter is designed and applied to the modeled focus curves. The experimental results demonstrate the effectiveness of proposed method.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130481932","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049930
Tanakorn Inthasuth, Poonpong Uarchoojitt, W. Boonsong, Natapon Kaewthong
One quality of a data communication system is supposed to consider the low latency delay time between a data sender and a receiver. Communication delays sometimes occur in a Narrowband Internet of things network (NB-IoT) between the end device (ED) and the cloud platform. The communication problem may be caused by packet loss or congestion in the system. Therefore, this paper measures transmission time between the ED and the cloud platform. The system was designed for the particulate matter 2.5 (PM2.5) monitoring system via the NB-IoT network. The system consists of the NB-IoT-based PM2.5 detector and real-time monitoring by the cloud platform. In the experiment, the test was conducted to document the data collection for 1 hour. The results showed that in the typical case, the transmission time between ED and cloud was an average of 2.01 seconds. However, in case of connection problems, the ED will retransmit until it succeeds, taking a maximum of 28.76 seconds Therefore, when configuring the period, it is necessary to consider transmission to minimize data loss, which is a limitation of communication. Namely, those problems will be solved and improved in future work to get higher performance with a low latency delay time of each round-trip time (RTT) of data communication.
{"title":"Time-Based Performance Analysis of Narrowband Internet of Things (NB-IoT) for Particulate Matter Monitoring System","authors":"Tanakorn Inthasuth, Poonpong Uarchoojitt, W. Boonsong, Natapon Kaewthong","doi":"10.1109/ICEIC57457.2023.10049930","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049930","url":null,"abstract":"One quality of a data communication system is supposed to consider the low latency delay time between a data sender and a receiver. Communication delays sometimes occur in a Narrowband Internet of things network (NB-IoT) between the end device (ED) and the cloud platform. The communication problem may be caused by packet loss or congestion in the system. Therefore, this paper measures transmission time between the ED and the cloud platform. The system was designed for the particulate matter 2.5 (PM2.5) monitoring system via the NB-IoT network. The system consists of the NB-IoT-based PM2.5 detector and real-time monitoring by the cloud platform. In the experiment, the test was conducted to document the data collection for 1 hour. The results showed that in the typical case, the transmission time between ED and cloud was an average of 2.01 seconds. However, in case of connection problems, the ED will retransmit until it succeeds, taking a maximum of 28.76 seconds Therefore, when configuring the period, it is necessary to consider transmission to minimize data loss, which is a limitation of communication. Namely, those problems will be solved and improved in future work to get higher performance with a low latency delay time of each round-trip time (RTT) of data communication.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130503296","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049857
Yujin Jeon, Eun-Gyeong Ham, Jaeyun Lim, Ji-Hoon Kim
AES-CCM (Advanced Encryption Standard-Counter with CBC-MAC) is an encryption and decryption method that guarantees the integrity and authenticity of data, commonly used for wireless security systems. The AES Core performs the central functions for encryption and decryption of the AES-CCM, so many studies are being conducted to implement it efficiently. In this paper, to investigate the impact of the hardware-software partitioning of the AES-CCM implementation for Bluetooth LE (Low Energy), we compare the results between the software-only implementation and hardware-software co-design approach where the AES encryption into the hardware and the rest into the software. As a result, a well-balanced co-design implementation of the AES-CCM shows a reduced latency of up to 77.4% compared to the software-only implementation for the Bluetooth LE system.
{"title":"Hardware-Software Co-Design of AES-CCM for Bluetooth LE Security","authors":"Yujin Jeon, Eun-Gyeong Ham, Jaeyun Lim, Ji-Hoon Kim","doi":"10.1109/ICEIC57457.2023.10049857","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049857","url":null,"abstract":"AES-CCM (Advanced Encryption Standard-Counter with CBC-MAC) is an encryption and decryption method that guarantees the integrity and authenticity of data, commonly used for wireless security systems. The AES Core performs the central functions for encryption and decryption of the AES-CCM, so many studies are being conducted to implement it efficiently. In this paper, to investigate the impact of the hardware-software partitioning of the AES-CCM implementation for Bluetooth LE (Low Energy), we compare the results between the software-only implementation and hardware-software co-design approach where the AES encryption into the hardware and the rest into the software. As a result, a well-balanced co-design implementation of the AES-CCM shows a reduced latency of up to 77.4% compared to the software-only implementation for the Bluetooth LE system.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129205082","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-02-05DOI: 10.1109/ICEIC57457.2023.10049899
Yeong-Jun Kim, Donggoo Kang, Yeongheon Mok, Sunkyu Kwon, J. Paik
Deep-learning based approach has solved various medical imaging problems successfully. Since the lack of training data issues caused by patient privacy, the few-shot learning method has been studied widely. However, this issue still afflicts model performance even in few-shot learning methods. To solve this issue, it is important to quickly optimize the initial parameter values using a small amount of data. In addition, to utilize small data effectively, it is important to design the objective function for segmentation suitable for GT (Ground Truth) with few-shots. In this paper, we experiment with various algorithms using the MAML (Model Agnostic Meta-Learning) method. And we propose an optimal few-shot semantic segmentation network. The proposed method uses a gradient descent algorithm and optimizer parameter decomposition method to ensure fast convergence with fewer data. Experimental results show high performance and fast convergence using fewer datasets than conventional methods.
{"title":"A Review on Few-shot Learning for Medical Image Segmentation","authors":"Yeong-Jun Kim, Donggoo Kang, Yeongheon Mok, Sunkyu Kwon, J. Paik","doi":"10.1109/ICEIC57457.2023.10049899","DOIUrl":"https://doi.org/10.1109/ICEIC57457.2023.10049899","url":null,"abstract":"Deep-learning based approach has solved various medical imaging problems successfully. Since the lack of training data issues caused by patient privacy, the few-shot learning method has been studied widely. However, this issue still afflicts model performance even in few-shot learning methods. To solve this issue, it is important to quickly optimize the initial parameter values using a small amount of data. In addition, to utilize small data effectively, it is important to design the objective function for segmentation suitable for GT (Ground Truth) with few-shots. In this paper, we experiment with various algorithms using the MAML (Model Agnostic Meta-Learning) method. And we propose an optimal few-shot semantic segmentation network. The proposed method uses a gradient descent algorithm and optimizer parameter decomposition method to ensure fast convergence with fewer data. Experimental results show high performance and fast convergence using fewer datasets than conventional methods.","PeriodicalId":373752,"journal":{"name":"2023 International Conference on Electronics, Information, and Communication (ICEIC)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128001404","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：