Visual inspection systems in industries have increasingly gained a lot of interests. Advances in manufacturing activities have led to mass production in order to reduce overall operational cost. The visual inspection systems provide instant quantitative feedback such as quantity and type of defects. In this paper, we present a visual inspection method of tiles industry using a deep learning approach. The deep learning approach is employed for segmenting cracks and backgrounds in tiles. Due to the small size of the cracks, image segmentation is crucial. Architecture for segmenting semantic objects in a color image is the main inspiration to be applied on this paper. Semantic segmentation is widely applied for image analysis in the real world, one of which is to conduct a visual inspection of tile surfaces where each pixel input of high-resolution images is categorized into a set of semantic labels. In order to test the performance of the segmentation algorithm, SegNet architecture with the DeepLabV3plus were compared. A new dataset named UBIN is also proposed as a training and evaluation data. The training data that we have collected shows promising results on visual inspection when using the proposed algorithm. We believe that this work could improve to a more advanced manufacturing industries.
{"title":"Tile Surface Segmentation Using Deep Convolutional Encoder-Decoder Architecture","authors":"Evianita Dewi Fajrianti, Endah Suryawati Ningrum, Anhar Risnumawan, Kerent Vidia Madalena","doi":"10.1109/IES50839.2020.9231575","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231575","url":null,"abstract":"Visual inspection systems in industries have increasingly gained a lot of interests. Advances in manufacturing activities have led to mass production in order to reduce overall operational cost. The visual inspection systems provide instant quantitative feedback such as quantity and type of defects. In this paper, we present a visual inspection method of tiles industry using a deep learning approach. The deep learning approach is employed for segmenting cracks and backgrounds in tiles. Due to the small size of the cracks, image segmentation is crucial. Architecture for segmenting semantic objects in a color image is the main inspiration to be applied on this paper. Semantic segmentation is widely applied for image analysis in the real world, one of which is to conduct a visual inspection of tile surfaces where each pixel input of high-resolution images is categorized into a set of semantic labels. In order to test the performance of the segmentation algorithm, SegNet architecture with the DeepLabV3plus were compared. A new dataset named UBIN is also proposed as a training and evaluation data. The training data that we have collected shows promising results on visual inspection when using the proposed algorithm. We believe that this work could improve to a more advanced manufacturing industries.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125175370","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231644
Muhammad Rizal Firmanda, Bima Sena Bayu Dewantara, R. Sigit
The availability of health check facilities that are increasingly affordable in terms of cost and distance is very useful for the community. Therefore, the presence of a healthcare kiosk that is installed everywhere will certainly be very beneficial for the wider community. In this paper, we developed a user-login method in the healthcare kiosk without utilizing any additional tools so that it is quite efficient, using face recognition biometric technology. We added the invariant illumination feature to face recognition technology to ensure that the healthcare kiosk system will continue to work even in places with changing lighting. This feature uses the light intensity contrast adjustment in the image automatically by employing the Fuzzy Inference System (FIS) and Genetic Algorithm (GA). Based on the results of testing the user-login system, we get an accuracy of 85.57% with an ideal distance of 30-60 cm. The system works with maximum performance in some experimental conditions such as normal lighting, backlighting, direct-lighting, low-lighting, and dark. Users that use facial recognition as a login method, the facial recognition program will capture the user's face and match it to the database. If the user is registered in the database, the system will inform the user that the user is logged in successfully, otherwise, if the user is not logged in, the system will notify the user as unknown.
{"title":"Implementation of Illumination Invariant Face Recognition for Accessing User Record in Healthcare Kiosk","authors":"Muhammad Rizal Firmanda, Bima Sena Bayu Dewantara, R. Sigit","doi":"10.1109/IES50839.2020.9231644","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231644","url":null,"abstract":"The availability of health check facilities that are increasingly affordable in terms of cost and distance is very useful for the community. Therefore, the presence of a healthcare kiosk that is installed everywhere will certainly be very beneficial for the wider community. In this paper, we developed a user-login method in the healthcare kiosk without utilizing any additional tools so that it is quite efficient, using face recognition biometric technology. We added the invariant illumination feature to face recognition technology to ensure that the healthcare kiosk system will continue to work even in places with changing lighting. This feature uses the light intensity contrast adjustment in the image automatically by employing the Fuzzy Inference System (FIS) and Genetic Algorithm (GA). Based on the results of testing the user-login system, we get an accuracy of 85.57% with an ideal distance of 30-60 cm. The system works with maximum performance in some experimental conditions such as normal lighting, backlighting, direct-lighting, low-lighting, and dark. Users that use facial recognition as a login method, the facial recognition program will capture the user's face and match it to the database. If the user is registered in the database, the system will inform the user that the user is logged in successfully, otherwise, if the user is not logged in, the system will notify the user as unknown.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130997665","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231815
D. O. Anggriawan, E. Wahjono, I. Sudiharto, Aji Akbar Firdaus, Dianing Novita Nurmala Putri, Anang Budikarso
This paper presents the proposed algorithms for identification of short duration voltage variations (SDVV) as voltage sag dan voltage swell. The proposed algorithms are short time fourier transform (STFT) and artificial neural network (ANN). STFT is used to signal analysis of SDVV. However, SDVV have characteristics of non-stationary signal, which it is not can be detected by fast fourier transform (FFT). Moreover, ANN is used to identification of SDVV, which it identifies five types of SDVV such as normal signal, voltage sag, voltage swell, voltage sag combined harmonic and voltage swell combined harmonic. Output STFT is used to ANN for identification. The simulation is conducted by STFT comparing of FFT. Whereas, to evaluate of ANN with variation of neurons. The simulation result show that STFT more accurate compared by FFT to detection of SDVV. Moreover, ANN has good accuracy for SDVV types identification, which ANN with 10 x 10 neurons in hidden layer has accuracy of 100 %.
{"title":"Identification of Short Duration Voltage Variations Based on Short Time Fourier Transform and Artificial Neural Network","authors":"D. O. Anggriawan, E. Wahjono, I. Sudiharto, Aji Akbar Firdaus, Dianing Novita Nurmala Putri, Anang Budikarso","doi":"10.1109/IES50839.2020.9231815","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231815","url":null,"abstract":"This paper presents the proposed algorithms for identification of short duration voltage variations (SDVV) as voltage sag dan voltage swell. The proposed algorithms are short time fourier transform (STFT) and artificial neural network (ANN). STFT is used to signal analysis of SDVV. However, SDVV have characteristics of non-stationary signal, which it is not can be detected by fast fourier transform (FFT). Moreover, ANN is used to identification of SDVV, which it identifies five types of SDVV such as normal signal, voltage sag, voltage swell, voltage sag combined harmonic and voltage swell combined harmonic. Output STFT is used to ANN for identification. The simulation is conducted by STFT comparing of FFT. Whereas, to evaluate of ANN with variation of neurons. The simulation result show that STFT more accurate compared by FFT to detection of SDVV. Moreover, ANN has good accuracy for SDVV types identification, which ANN with 10 x 10 neurons in hidden layer has accuracy of 100 %.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132791030","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231890
Fitriyah Fitriyah, M. Z. Efendi, Farid Dwi Murdianto
The increasing development of renewable energy, especially photovoltaic (PV) more applicable in daily use. Unfortunately, PV has the disadvantage that it is vulnerable to shadow exposures that decrease power output depending on the scale of the shadow. The disadvantage includes the shadow of buildings, leaves, trees, etc. Shaded PV surface has two peak power conditions named Global Maximum Power Point (GMPP) and Local Maximum Power Point (LMPP). These conditions cause MPPT to be trapped in the LMPP so that the power obtained is not the actual power. The conventional method can be trapped in LMPP because it cannot distinguish GMPP and LMPP. These problems can be solved by using the Human Psychology Optimization (HPO) algorithm. This algorithm was chosen to overcome the effects of partial shading conditions so that MPPT can reach GMPP without getting stuck in LMPP. This algorithm is connected to ZETA Converter to produce real maximum power points. This research uses four shading patterns with different irradiation. HPO algorithm achieves the highest accuracy 99.99% with a tracking time 0.326 seconds occurring in the first shading pattern.
{"title":"Modeling and Simulation of MPPT ZETA Converter Using Human Psychology Optimization Algorithm Under Partial Shading Condition","authors":"Fitriyah Fitriyah, M. Z. Efendi, Farid Dwi Murdianto","doi":"10.1109/IES50839.2020.9231890","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231890","url":null,"abstract":"The increasing development of renewable energy, especially photovoltaic (PV) more applicable in daily use. Unfortunately, PV has the disadvantage that it is vulnerable to shadow exposures that decrease power output depending on the scale of the shadow. The disadvantage includes the shadow of buildings, leaves, trees, etc. Shaded PV surface has two peak power conditions named Global Maximum Power Point (GMPP) and Local Maximum Power Point (LMPP). These conditions cause MPPT to be trapped in the LMPP so that the power obtained is not the actual power. The conventional method can be trapped in LMPP because it cannot distinguish GMPP and LMPP. These problems can be solved by using the Human Psychology Optimization (HPO) algorithm. This algorithm was chosen to overcome the effects of partial shading conditions so that MPPT can reach GMPP without getting stuck in LMPP. This algorithm is connected to ZETA Converter to produce real maximum power points. This research uses four shading patterns with different irradiation. HPO algorithm achieves the highest accuracy 99.99% with a tracking time 0.326 seconds occurring in the first shading pattern.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115376526","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231640
M. T. Sumadi, Mike Yuliana, Amang Sudarsono
In symmetric cryptography systems have problems in the distribution of secret keys. The two users who will communicate require sharing keys through the public channel. The proposed solution to overcome these problems is to utilize information from the physical layer (e.g. RSS). Received Signal Strength (RSS) is an indicator for measuring the power received by wireless devices. The advantage of secret key extraction using physical layer information from a wireless channel is that it allows both devices within the transmission range to extract the secret key together. In this paper, we propose a secret key generation scheme adopted from an existing scheme with modifications to improve performance. Our proposed system is applied to static and dynamic conditions to test performance. The proposed algorithm is able to obtain a reduction in KDR (Key Disagreement Rate) up to 48.42% and an increase in the KGR (Key Generation Rate) up to 23.35% when compared to the existing scheme. Our proposed system also successfully passed the randomness using the NIST test with the approximate value of entropy generated 0.80 in static conditions and 0.81 in dynamic conditions.
{"title":"Performance Improvement Based on Modified Lossless Quantization (MLQ) for Secret Key Generation Extracted from Received Signal Strength","authors":"M. T. Sumadi, Mike Yuliana, Amang Sudarsono","doi":"10.1109/IES50839.2020.9231640","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231640","url":null,"abstract":"In symmetric cryptography systems have problems in the distribution of secret keys. The two users who will communicate require sharing keys through the public channel. The proposed solution to overcome these problems is to utilize information from the physical layer (e.g. RSS). Received Signal Strength (RSS) is an indicator for measuring the power received by wireless devices. The advantage of secret key extraction using physical layer information from a wireless channel is that it allows both devices within the transmission range to extract the secret key together. In this paper, we propose a secret key generation scheme adopted from an existing scheme with modifications to improve performance. Our proposed system is applied to static and dynamic conditions to test performance. The proposed algorithm is able to obtain a reduction in KDR (Key Disagreement Rate) up to 48.42% and an increase in the KGR (Key Generation Rate) up to 23.35% when compared to the existing scheme. Our proposed system also successfully passed the randomness using the NIST test with the approximate value of entropy generated 0.80 in static conditions and 0.81 in dynamic conditions.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"248 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114247630","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231728
A. Fahmi, D. Purwitasari, S. Sumpeno, M. Purnomo
Dengue fever caused by dengue virus infection is a severe health threat that can lead to death. In the medical and health field, to classify data, data mining exploitation and classification methods have an essential role in predicting disease. Two main criteria are crucial to diagnosing dengue virus infection, namely the criteria clinical diagnosis and laboratory diagnosis. Dengue infection based on clinical signs and symptoms, as well as laboratory examinations, is made in three clinical diagnosis criteria, which consist of dengue fever (DF), dengue hemorrhagic fever (DHF), and dengue shock syndrome (DSS). This study was conducted with the primary objective to test and evaluate eight different classification algorithms to find the best algorithm in terms of efficiency and effectiveness. Classification algorithm used to predict dengue virus infection cases into three classes of DF, DHF, and DSS based on the performance of accuracy, precision, and recall. The classification algorithm used in this comparison were Neural Networks (NN), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Decision Tree, Random Forest, Naïve Bayes, AdaBoost, and Logistic Regression. The dataset called DBDDKK was collected from the Division of Disease Prevention and Control in the Semarang City Health Office, Central Java, Indonesia. Impute missing values, selection relevant feature, and normalize feature conducted in the preprocessing stage resulted in 14,019 records with 16 attributes for each record. Then the data were split into 70% for training data and 30% for testing data. Cross-validation with the number of folds 10 is applied to validate the accuracy during the dataset training process. The result of the comparison shows that the NN algorithm has the best accuracy that was over other algorithms.
{"title":"Performance Evaluation of Classifiers for Predicting Infection Cases of Dengue Virus Based on Clinical Diagnosis Criteria","authors":"A. Fahmi, D. Purwitasari, S. Sumpeno, M. Purnomo","doi":"10.1109/IES50839.2020.9231728","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231728","url":null,"abstract":"Dengue fever caused by dengue virus infection is a severe health threat that can lead to death. In the medical and health field, to classify data, data mining exploitation and classification methods have an essential role in predicting disease. Two main criteria are crucial to diagnosing dengue virus infection, namely the criteria clinical diagnosis and laboratory diagnosis. Dengue infection based on clinical signs and symptoms, as well as laboratory examinations, is made in three clinical diagnosis criteria, which consist of dengue fever (DF), dengue hemorrhagic fever (DHF), and dengue shock syndrome (DSS). This study was conducted with the primary objective to test and evaluate eight different classification algorithms to find the best algorithm in terms of efficiency and effectiveness. Classification algorithm used to predict dengue virus infection cases into three classes of DF, DHF, and DSS based on the performance of accuracy, precision, and recall. The classification algorithm used in this comparison were Neural Networks (NN), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Decision Tree, Random Forest, Naïve Bayes, AdaBoost, and Logistic Regression. The dataset called DBDDKK was collected from the Division of Disease Prevention and Control in the Semarang City Health Office, Central Java, Indonesia. Impute missing values, selection relevant feature, and normalize feature conducted in the preprocessing stage resulted in 14,019 records with 16 attributes for each record. Then the data were split into 70% for training data and 30% for testing data. Cross-validation with the number of folds 10 is applied to validate the accuracy during the dataset training process. The result of the comparison shows that the NN algorithm has the best accuracy that was over other algorithms.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121130999","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231781
G. Pongthanisorn, W. Viriyavit, T. Prakayapan, S. Deepaisam, V. Somlertlamvanich
The Internet of Things (IoT) has become more practical nowadays. Various IoT applications are now being developed and deployed in order to attenuate our daily life problems. The global society is currently suffering problems associate to different aspects and ones that require our concern are regarding the aging society. Many and more countries are entering their aging society era while their healthcare sections have yet effective solutions to overcome the corresponding problems. Therefore, we propose the Elder Care System (ECS) for monitoring behaviours of elderly patients on the bed equipped with our designed system. The system includes notification system, in-bed position prediction system and real-time monitoring system. This paper demonstrates equipment, system architecture and dataflow. The result of our system deployment is discussed.
{"title":"ECS: Elderly Care System for Fall and Bedsore Prevention using Non-Constraint Sensor","authors":"G. Pongthanisorn, W. Viriyavit, T. Prakayapan, S. Deepaisam, V. Somlertlamvanich","doi":"10.1109/IES50839.2020.9231781","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231781","url":null,"abstract":"The Internet of Things (IoT) has become more practical nowadays. Various IoT applications are now being developed and deployed in order to attenuate our daily life problems. The global society is currently suffering problems associate to different aspects and ones that require our concern are regarding the aging society. Many and more countries are entering their aging society era while their healthcare sections have yet effective solutions to overcome the corresponding problems. Therefore, we propose the Elder Care System (ECS) for monitoring behaviours of elderly patients on the bed equipped with our designed system. The system includes notification system, in-bed position prediction system and real-time monitoring system. This paper demonstrates equipment, system architecture and dataflow. The result of our system deployment is discussed.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122371805","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231813
Rasyiq Farandi, Amang Sudarsono, M. Z. S. Hadi
This research focuses on tracking user-behavior based on applications usage of the user’s smartphone and creates a conclusion of user behavior when he/she has a trip of traveling. To track the application usage from the user’s smartphone, a library from Android SDK is utilized in creating an instruction that can track the application usage from the user’s smartphone. While he/she is on a trip, his/her positions can be tracked by using Google Maps API as well integrated on the application. To realize these functionalities, a heuristic method is adopted. In this application, user behavior data that have been made are stored in the database. The data sent are secured using the RSA algorithm and equipped with message digest verification. The results show that our proposed application is worked properly within the testing scope for behavioral generation during traveling.
{"title":"A Mobile Application of User Traveling-Behavior Tracking using Heuristic Method","authors":"Rasyiq Farandi, Amang Sudarsono, M. Z. S. Hadi","doi":"10.1109/IES50839.2020.9231813","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231813","url":null,"abstract":"This research focuses on tracking user-behavior based on applications usage of the user’s smartphone and creates a conclusion of user behavior when he/she has a trip of traveling. To track the application usage from the user’s smartphone, a library from Android SDK is utilized in creating an instruction that can track the application usage from the user’s smartphone. While he/she is on a trip, his/her positions can be tracked by using Google Maps API as well integrated on the application. To realize these functionalities, a heuristic method is adopted. In this application, user behavior data that have been made are stored in the database. The data sent are secured using the RSA algorithm and equipped with message digest verification. The results show that our proposed application is worked properly within the testing scope for behavioral generation during traveling.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124727069","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231950
Berlian Rahmy Lidiawaty, Mohammad Isa Irawan, R. V. Hari Ginardi
Sellers in online marketplace who use batik Solo’s pattern name as their product title, but mistaken it with product image that doesn’t contain the pattern, could causes the buyer to wear batik Solo in wrong event. Every pattern in batik Solo has crucial meaning and specialize to be worn in different event such as wedding ceremony or funeral. Therefore this research has the purpose to build a system that can verify batik Solo’s product image depending on the batik pattern name that the seller wants to use as their product title in online marketplace. First, the research input batik Solo pattern name in the four biggest online marketplaces in Indonesia, which are Tokopedia, Bukalapak, Shopee and Lazada. Second, it scrapped some of the images result from every marketplace. Then the research makes a system that can identify the image’s texture from marketplace and identify the image’s texture from the data set that was prepared before. This process could be completed using the SURF method. Next, an image from an online marketplace that contains a specific pattern will input to the system to find if it retrieves image from data set’s image with a similar intended pattern or not. The system verification will label an image as true if it can retrieve some images from data set, and label an image as false if it can’t retrieve image from data set. The output label from the system will be compared with the label from human judgement to measure the accuracy of the system. The results are, the highest accuracy is 74.76% and the highest recall is 94.55%.
{"title":"Image Pattern Verification Based On Seller's Batik Solo Product Name Using SURF As A Texture Based Image Retrieval","authors":"Berlian Rahmy Lidiawaty, Mohammad Isa Irawan, R. V. Hari Ginardi","doi":"10.1109/IES50839.2020.9231950","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231950","url":null,"abstract":"Sellers in online marketplace who use batik Solo’s pattern name as their product title, but mistaken it with product image that doesn’t contain the pattern, could causes the buyer to wear batik Solo in wrong event. Every pattern in batik Solo has crucial meaning and specialize to be worn in different event such as wedding ceremony or funeral. Therefore this research has the purpose to build a system that can verify batik Solo’s product image depending on the batik pattern name that the seller wants to use as their product title in online marketplace. First, the research input batik Solo pattern name in the four biggest online marketplaces in Indonesia, which are Tokopedia, Bukalapak, Shopee and Lazada. Second, it scrapped some of the images result from every marketplace. Then the research makes a system that can identify the image’s texture from marketplace and identify the image’s texture from the data set that was prepared before. This process could be completed using the SURF method. Next, an image from an online marketplace that contains a specific pattern will input to the system to find if it retrieves image from data set’s image with a similar intended pattern or not. The system verification will label an image as true if it can retrieve some images from data set, and label an image as false if it can’t retrieve image from data set. The output label from the system will be compared with the label from human judgement to measure the accuracy of the system. The results are, the highest accuracy is 74.76% and the highest recall is 94.55%.","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122123971","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 : 2020-09-01DOI: 10.1109/IES50839.2020.9231952
Mufidatun Nisa Nur Lailiyah, A. Basofi, A. Fariza
In Indonesia, age identity plays an important role in deciding many things, for example, to determine the level of education, medical treatment, and health, to determine the age allowed to get married, to get a job, etc. An effective way to overcome age counterfeiting is to recognize facial images that have unique biometric features. Age development is generally indicated by skin texture and facial structure, this makes it quite difficult to estimate age. Therefore, we need automatic age identification that can be convinced, can be accounted for in the public interest, and has a high closeness. This paper proposed the age estimation of Indonesian face using Deep Convolutional Neural Networks (CNN) DenseNet-161 model architecture approach. The dataset is collected with a range of 7-22 years old of 2300 face image of Indonesian. We compare the prediction result with the custom architecture of CNN with 3 convolution layer and 3 fully connected. The prediction results of the DenseNet-161 model achieved very good prediction results (MAE = 3.02, Accuracy = 67.93%, and R-Squared = 0.99) than the custom model (MAE = 3.17, Accuracy = 64.47%, and R-Squared = 0.97).
{"title":"Age Estimation Based on Indonesian Face Recognition using Convolutional Neural Networks","authors":"Mufidatun Nisa Nur Lailiyah, A. Basofi, A. Fariza","doi":"10.1109/IES50839.2020.9231952","DOIUrl":"https://doi.org/10.1109/IES50839.2020.9231952","url":null,"abstract":"In Indonesia, age identity plays an important role in deciding many things, for example, to determine the level of education, medical treatment, and health, to determine the age allowed to get married, to get a job, etc. An effective way to overcome age counterfeiting is to recognize facial images that have unique biometric features. Age development is generally indicated by skin texture and facial structure, this makes it quite difficult to estimate age. Therefore, we need automatic age identification that can be convinced, can be accounted for in the public interest, and has a high closeness. This paper proposed the age estimation of Indonesian face using Deep Convolutional Neural Networks (CNN) DenseNet-161 model architecture approach. The dataset is collected with a range of 7-22 years old of 2300 face image of Indonesian. We compare the prediction result with the custom architecture of CNN with 3 convolution layer and 3 fully connected. The prediction results of the DenseNet-161 model achieved very good prediction results (MAE = 3.02, Accuracy = 67.93%, and R-Squared = 0.99) than the custom model (MAE = 3.17, Accuracy = 64.47%, and R-Squared = 0.97).","PeriodicalId":344685,"journal":{"name":"2020 International Electronics Symposium (IES)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127784012","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：