Pub Date : 2018-05-07DOI: 10.1109/ISACV.2018.8354044
Nazmul Hussain, Hai H. Wang, C. Buckingham
Autonomic computing was intended to tackle the growing complexity of Information Technology infrastructure by making it self-managing and self-adaptive. The core idea is to endow the system with enough intelligence to monitor continuously all aspects of the changing environments and resources, and to control management decisions according to high-level policies. For several years, great efforts have been devoted to the study of system performance, security, and fault management issues, but without much attention paid to self-adaptive social-collaborative system development. This may be because it is difficult to create such autonomic systems, which must sense and adapt to ongoing social context changes and support cyber-physical collaborations with minimal human involvement. These collaborations will have interactions between human and non-human entities that need to be self-managing with adaptive goals. This paper tackles the problem by introducing a new Generic Autonomic Social-Collaborative Framework (GASCF). It focuses on a high-level social-context based self-adaptive system, and its use of intelligent agents called autonomic adapters(AAs) that are driven by predefined policies. The paper describes the architecture of autonomic adapters and the general representation of policies. It explores the effectiveness of the approach by applying it to a large-scale collaborative healthcare service called GRaCE (https://www.egrist.org/) that supports mental-health within the United Kingdom National Health Service and other organisations.
{"title":"Policy based generic autonomic adapter for a context-aware social-collaborative system","authors":"Nazmul Hussain, Hai H. Wang, C. Buckingham","doi":"10.1109/ISACV.2018.8354044","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354044","url":null,"abstract":"Autonomic computing was intended to tackle the growing complexity of Information Technology infrastructure by making it self-managing and self-adaptive. The core idea is to endow the system with enough intelligence to monitor continuously all aspects of the changing environments and resources, and to control management decisions according to high-level policies. For several years, great efforts have been devoted to the study of system performance, security, and fault management issues, but without much attention paid to self-adaptive social-collaborative system development. This may be because it is difficult to create such autonomic systems, which must sense and adapt to ongoing social context changes and support cyber-physical collaborations with minimal human involvement. These collaborations will have interactions between human and non-human entities that need to be self-managing with adaptive goals. This paper tackles the problem by introducing a new Generic Autonomic Social-Collaborative Framework (GASCF). It focuses on a high-level social-context based self-adaptive system, and its use of intelligent agents called autonomic adapters(AAs) that are driven by predefined policies. The paper describes the architecture of autonomic adapters and the general representation of policies. It explores the effectiveness of the approach by applying it to a large-scale collaborative healthcare service called GRaCE (https://www.egrist.org/) that supports mental-health within the United Kingdom National Health Service and other organisations.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116825537","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 : 2018-05-03DOI: 10.1109/ISACV.2018.8354068
Ronald Saeijs, W. E. Tjon a Ten, P. D. De with
This paper presents a new algorithm for dual-camera 3D head tracking, intended for clinical infant monitoring. The paper includes a brief motivation with reference to the state-of-the-art in face-related image analysis. The proposed algorithm uses a clipped-ellipsoid head model and 3D head pose recovery by joint alignment of paired templates based on dense-HOG features. In the algorithm, template pairs are dynamically extracted and a limited number of template pairs are stored and re-used for drift reduction. We report experimental results on real-life videos of infants in bed in a hospital, captured in visual light as well as near-infrared light. Results show consistently good tracking behavior. For challenging video sequences, the mean tracking error in terms of endocanthion location error relative to the innercanthal distance remains below 30%. This error has proven to be sufficiently low for 3D head tracking to support infant face analysis. For this reason, the proposed algorithm is used successfully in an infant monitoring system under development.
{"title":"Dual-camera 3D head tracking for clinical infant monitoring","authors":"Ronald Saeijs, W. E. Tjon a Ten, P. D. De with","doi":"10.1109/ISACV.2018.8354068","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354068","url":null,"abstract":"This paper presents a new algorithm for dual-camera 3D head tracking, intended for clinical infant monitoring. The paper includes a brief motivation with reference to the state-of-the-art in face-related image analysis. The proposed algorithm uses a clipped-ellipsoid head model and 3D head pose recovery by joint alignment of paired templates based on dense-HOG features. In the algorithm, template pairs are dynamically extracted and a limited number of template pairs are stored and re-used for drift reduction. We report experimental results on real-life videos of infants in bed in a hospital, captured in visual light as well as near-infrared light. Results show consistently good tracking behavior. For challenging video sequences, the mean tracking error in terms of endocanthion location error relative to the innercanthal distance remains below 30%. This error has proven to be sufficiently low for 3D head tracking to support infant face analysis. For this reason, the proposed algorithm is used successfully in an infant monitoring system under development.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-05-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130479276","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 : 2018-04-02DOI: 10.1109/ISACV.2018.8354024
H. Elaidi, Younes Elhaddar, Zahra Benabbou, Hassan Abbar
Clustering is a technique which is commonly known in the domain of machine learning as an unsupervised method, it aims at constructing from a set of objects some different groups which are as homogeneous as possible. On the other hand support vector machines (SVM) and binary decision trees (BDT) were proposed and developed as supervised learning techniques where the output assembly is previously known. In this work we will try to build a clustering algorithm that uses the two supervised methods we cited above.
{"title":"An idea of a clustering algorithm using support vector machines based on binary decision tree","authors":"H. Elaidi, Younes Elhaddar, Zahra Benabbou, Hassan Abbar","doi":"10.1109/ISACV.2018.8354024","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354024","url":null,"abstract":"Clustering is a technique which is commonly known in the domain of machine learning as an unsupervised method, it aims at constructing from a set of objects some different groups which are as homogeneous as possible. On the other hand support vector machines (SVM) and binary decision trees (BDT) were proposed and developed as supervised learning techniques where the output assembly is previously known. In this work we will try to build a clustering algorithm that uses the two supervised methods we cited above.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"525 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122918662","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 : 2018-04-02DOI: 10.1109/ISACV.2018.8354022
N. Chater, T. Mazri, M. Benbrahim
This paper presents characteristics improvement of patch array antenna used for electronic scanning radar application. This basic array antenna, designed on low cost substrate FR-4 and fed by microstrip technique, consists of 8 patches and operates at frequency 3GHz. However, this structure has two disadvantages a low gain value by dint of using a lossy material FR4 as a substrate, and a narrow bandwidth which is due to microstrip antenna limitations. Hence, the objective of this work is to increase the gain and the bandwidth of this structure. Therefore two techniques will be proposed separately in this paper: parasitic patches and multi layer substrate. For the first one, it consists of adding a defined number of parasitic patches, the distance between driven and parasitic patches will be evaluated to ensure strong coupling between them. For the second technique, a second layer of FR4 coated with an annealed copper of 0.035mm is added to the substrate. An air gap of 0.04x is used to separate between both FR-4 layers. The design and simulation of the array antenna and the modified structures will be performed using Advanced Design System (ADS) software. Simulation results of each technique will be compared first to basic array antenna simulation results and compared thereafter to each other.
{"title":"Array antenna characteristics enhancement for electronic scanning radar application: Parasitic patches and multi layer techniques","authors":"N. Chater, T. Mazri, M. Benbrahim","doi":"10.1109/ISACV.2018.8354022","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354022","url":null,"abstract":"This paper presents characteristics improvement of patch array antenna used for electronic scanning radar application. This basic array antenna, designed on low cost substrate FR-4 and fed by microstrip technique, consists of 8 patches and operates at frequency 3GHz. However, this structure has two disadvantages a low gain value by dint of using a lossy material FR4 as a substrate, and a narrow bandwidth which is due to microstrip antenna limitations. Hence, the objective of this work is to increase the gain and the bandwidth of this structure. Therefore two techniques will be proposed separately in this paper: parasitic patches and multi layer substrate. For the first one, it consists of adding a defined number of parasitic patches, the distance between driven and parasitic patches will be evaluated to ensure strong coupling between them. For the second technique, a second layer of FR4 coated with an annealed copper of 0.035mm is added to the substrate. An air gap of 0.04x is used to separate between both FR-4 layers. The design and simulation of the array antenna and the modified structures will be performed using Advanced Design System (ADS) software. Simulation results of each technique will be compared first to basic array antenna simulation results and compared thereafter to each other.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127891340","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 : 2018-04-02DOI: 10.1109/ISACV.2018.8354014
Aziz Elfatimi, S. Bri, Adil Saadi
In this electronic paper, we will present the design and numerical simulation of a compact single-layer millimeter wave antenna for the future fifth generation (5G applications), excited by a microstrip line. Its design is based on a rectangular patch with notches of different geometries mounted on a dielectric substrate. The proposed antenna operates at two different resonance frequencies, 28.00 GHz and 38.00 GHz. The numerical resolution of the maxwell equations using the finite element method (FEM), shows that the antenna can provide two bandwidths around 921 MHz (3.29%) centered at 28.01 GHz and 1.0451 GHz (3)., 72%) centered at 38.03 MHz. GHz. The reflection coefficients are −23.8112 dB for (fr)i = 28.01 GHz and −17.0898 dB for (fr)2 = 38.03 GHz. The gain of the antenna varies from 8.0527 dB for the band below 8.2869 dB for the upper band.
{"title":"Single feed compact millimeter wave antenna for future 5G applications","authors":"Aziz Elfatimi, S. Bri, Adil Saadi","doi":"10.1109/ISACV.2018.8354014","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354014","url":null,"abstract":"In this electronic paper, we will present the design and numerical simulation of a compact single-layer millimeter wave antenna for the future fifth generation (5G applications), excited by a microstrip line. Its design is based on a rectangular patch with notches of different geometries mounted on a dielectric substrate. The proposed antenna operates at two different resonance frequencies, 28.00 GHz and 38.00 GHz. The numerical resolution of the maxwell equations using the finite element method (FEM), shows that the antenna can provide two bandwidths around 921 MHz (3.29%) centered at 28.01 GHz and 1.0451 GHz (3)., 72%) centered at 38.03 MHz. GHz. The reflection coefficients are −23.8112 dB for (fr)i = 28.01 GHz and −17.0898 dB for (fr)2 = 38.03 GHz. The gain of the antenna varies from 8.0527 dB for the band below 8.2869 dB for the upper band.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"30 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131086779","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 : 2018-04-02DOI: 10.1109/ISACV.2018.8354012
Imad Lamouik, Ali Yahyaouy, M. A. Sabri
Traffic grids have always suffered from a lack of dynamic routing and path planning algorithms and relied only on static characteristics of the roads like the number of lanes, distance and speed limits to avoid and resolve traffic congestion, by routing traffic to a lighter traffic path. However, with the increased number of vehicles in urban areas these algorithms may have reached their limitation due to the huge increase in the state space in a limited computing power and memory environment. In this research we will introduce a dynamic routing system for traffic in intersections based on real-time traffic conditions such as individual vehicle speed, destination and traffic light status to provide the fasted path between a source and a target point. This system will exploit the recent advancements in the field of machine learning by leveraging the power of deep learning especially deep convolutional neural networks. Simulation shows that the proposed model results in a path that are generally fast and avoids frequent red light stops.
{"title":"Deep neural network dynamic traffic routing system for vehicles","authors":"Imad Lamouik, Ali Yahyaouy, M. A. Sabri","doi":"10.1109/ISACV.2018.8354012","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354012","url":null,"abstract":"Traffic grids have always suffered from a lack of dynamic routing and path planning algorithms and relied only on static characteristics of the roads like the number of lanes, distance and speed limits to avoid and resolve traffic congestion, by routing traffic to a lighter traffic path. However, with the increased number of vehicles in urban areas these algorithms may have reached their limitation due to the huge increase in the state space in a limited computing power and memory environment. In this research we will introduce a dynamic routing system for traffic in intersections based on real-time traffic conditions such as individual vehicle speed, destination and traffic light status to provide the fasted path between a source and a target point. This system will exploit the recent advancements in the field of machine learning by leveraging the power of deep learning especially deep convolutional neural networks. Simulation shows that the proposed model results in a path that are generally fast and avoids frequent red light stops.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"115 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117126910","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 : 2018-04-02DOI: 10.1109/ISACV.2018.8354082
Marouane Benmoussa, A. Mahmoudi
Human Computer Interaction received a great deal of attention this last decade. Last researches has turned to more natural interaction systems like gestural human machine interfaces. Recent works are attempting to solve the problem of hand gestures recognition using machine learning methods. Some of them are pretending to achieve very high performance. However, few of them are taking into account mandatory requirements to apply the workflow of a learning model, mainly data unbalance, model selection and generalization performance metric choice. In this work, we proposed a machine learning method for real time recognition of 16 gestures of user hands using the Kinect sensor that respects such requirements. The recognition is triggered only when there is a moving hand gesture. The method is based on the training of a Support Vector Machine model on hand depth data from which bag of words of SIFT and SURF descriptors are extracted. The data was kept balanced and the model kernel and parameters were selected using cross validation procedure. The method achieved 98% overall performance using the area under the ROC curve measure.
{"title":"Machine learning for hand gesture recognition using bag-of-words","authors":"Marouane Benmoussa, A. Mahmoudi","doi":"10.1109/ISACV.2018.8354082","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354082","url":null,"abstract":"Human Computer Interaction received a great deal of attention this last decade. Last researches has turned to more natural interaction systems like gestural human machine interfaces. Recent works are attempting to solve the problem of hand gestures recognition using machine learning methods. Some of them are pretending to achieve very high performance. However, few of them are taking into account mandatory requirements to apply the workflow of a learning model, mainly data unbalance, model selection and generalization performance metric choice. In this work, we proposed a machine learning method for real time recognition of 16 gestures of user hands using the Kinect sensor that respects such requirements. The recognition is triggered only when there is a moving hand gesture. The method is based on the training of a Support Vector Machine model on hand depth data from which bag of words of SIFT and SURF descriptors are extracted. The data was kept balanced and the model kernel and parameters were selected using cross validation procedure. The method achieved 98% overall performance using the area under the ROC curve measure.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131891368","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 : 2018-04-02DOI: 10.1109/ISACV.2018.8354019
Y. Tiamaz, N. Souissi
There are several procedures for applying Lean approach in order to improve business processes in different sectors. However, these procedures seem difficult to apprehend by the practitioners and do not allow them to realize all the desired objectives through business process improvement. This paper highlights the need to analyze and classify Lean implementation procedures in the literature to provide business process managers with the trends in procedures to implement Lean and improve business processes. In order to achieve this goal, a review of literature of Lean implementation procedures is established, then a classification of the procedures selected is proposed and finally, an analysis of these selected procedures is elaborated. This paper concludes that Tool-oriented procedures and hybrid procedures are the most adopted in all sectors and hybrid procedures that rely on tools, principles, and objectives contribute significantly to achieving the goals targeted by the business process manager.
{"title":"Classification of the lean implementation procedures for improving the business processes","authors":"Y. Tiamaz, N. Souissi","doi":"10.1109/ISACV.2018.8354019","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354019","url":null,"abstract":"There are several procedures for applying Lean approach in order to improve business processes in different sectors. However, these procedures seem difficult to apprehend by the practitioners and do not allow them to realize all the desired objectives through business process improvement. This paper highlights the need to analyze and classify Lean implementation procedures in the literature to provide business process managers with the trends in procedures to implement Lean and improve business processes. In order to achieve this goal, a review of literature of Lean implementation procedures is established, then a classification of the procedures selected is proposed and finally, an analysis of these selected procedures is elaborated. This paper concludes that Tool-oriented procedures and hybrid procedures are the most adopted in all sectors and hybrid procedures that rely on tools, principles, and objectives contribute significantly to achieving the goals targeted by the business process manager.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132077946","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 : 2018-04-02DOI: 10.1109/ISACV.2018.8354062
Mohamed Boudchiche, A. Mazroui
In this paper, we present an Arabic root extraction system. It provides the root of each word of a given sentence. It is an indispensable tool for several natural language processing applications such as search engines, text classification and information retrieval. The method of extraction used in this work runs in two steps. The first one consists in seeking of all the possible roots of each word analyzed out of context with the morphological analyzer Alkhalil Morpho Sys 2. Then, we develop in the second step a disambiguation approach based on continuous quadratic splines to choose among these roots the one that corresponds to the word context. We test this method on a representative corpus, and we obtained encouraging results with an accuracy of the order of 96%.
{"title":"Improving the Arabic root extraction by using the quadratic splines","authors":"Mohamed Boudchiche, A. Mazroui","doi":"10.1109/ISACV.2018.8354062","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354062","url":null,"abstract":"In this paper, we present an Arabic root extraction system. It provides the root of each word of a given sentence. It is an indispensable tool for several natural language processing applications such as search engines, text classification and information retrieval. The method of extraction used in this work runs in two steps. The first one consists in seeking of all the possible roots of each word analyzed out of context with the morphological analyzer Alkhalil Morpho Sys 2. Then, we develop in the second step a disambiguation approach based on continuous quadratic splines to choose among these roots the one that corresponds to the word context. We test this method on a representative corpus, and we obtained encouraging results with an accuracy of the order of 96%.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126357105","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 : 2018-04-02DOI: 10.1109/ISACV.2018.8354080
Achraf Oussidi, Azeddine Elhassouny
Generative models have found their way to the forefront of deep learning the last decade and so far, it seems that the hype will not fade away any time soon. In this paper, we give an overview of the most important building blocks of most recent revolutionary deep generative models such as RBM, DBM, DBN, VAE and GAN. We will also take a look at three of state-of-the-art generative models, namely PixelRNN, DRAW and NADE. We will delve into their unique architectures, the learning procedures and their potential and limitations. We will also review some of the known issues that arise when trying to design and train deep generative architectures using shallow ones and how different models deal with these issues. This paper is not meant to be a comprehensive study of these models, but rather a starting point for those who bear an interest in the field.
{"title":"Deep generative models: Survey","authors":"Achraf Oussidi, Azeddine Elhassouny","doi":"10.1109/ISACV.2018.8354080","DOIUrl":"https://doi.org/10.1109/ISACV.2018.8354080","url":null,"abstract":"Generative models have found their way to the forefront of deep learning the last decade and so far, it seems that the hype will not fade away any time soon. In this paper, we give an overview of the most important building blocks of most recent revolutionary deep generative models such as RBM, DBM, DBN, VAE and GAN. We will also take a look at three of state-of-the-art generative models, namely PixelRNN, DRAW and NADE. We will delve into their unique architectures, the learning procedures and their potential and limitations. We will also review some of the known issues that arise when trying to design and train deep generative architectures using shallow ones and how different models deal with these issues. This paper is not meant to be a comprehensive study of these models, but rather a starting point for those who bear an interest in the field.","PeriodicalId":184662,"journal":{"name":"2018 International Conference on Intelligent Systems and Computer Vision (ISCV)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115342932","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：