Pub Date : 2018-09-01DOI: 10.1109/ICCP.2018.8516648
Cristian Pintea, Eugen Pintea, Marcel Antal, Claudia Pop, T. Cioara, I. Anghel, I. Salomie
This paper addresses the problem of Data Centers (DCs) energy efficiency from a thermal perspective by extending the CloudSim framework to allow simulation and testing of thermal aware resource allocation policies aiming to minimize the cooling system energy consumption. The proposed framework, CoolCloudSim, can be used to develop and test new thermal aware Virtual Machine (VM) allocation strategies aiming to optimize the energy consumption of both cooling system and IT resources while meeting Service Level Agreements (SLAs). The default CloudSim architecture is extended by adding classes which contain mathematical models of the thermal processes within the server room. Furthermore, four new VM allocation policies that consider the cooling system energy consumption are developed based on the thermal and cooling system models. Finally, experiments are run to evaluate various metrics on a set of default CloudSim allocation algorithms and the proposed allocation algorithms. The results show that the proposed algorithms outperform the default CloudSim allocation strategy, Power Aware Best-Fit Decreasing (PABFD), in terms of overall energy consumption and the number of VM migrations, and have on average better results than other existing allocation strategies.
{"title":"CoolCloudSim: Integrating Cooling System Models in CloudSim","authors":"Cristian Pintea, Eugen Pintea, Marcel Antal, Claudia Pop, T. Cioara, I. Anghel, I. Salomie","doi":"10.1109/ICCP.2018.8516648","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516648","url":null,"abstract":"This paper addresses the problem of Data Centers (DCs) energy efficiency from a thermal perspective by extending the CloudSim framework to allow simulation and testing of thermal aware resource allocation policies aiming to minimize the cooling system energy consumption. The proposed framework, CoolCloudSim, can be used to develop and test new thermal aware Virtual Machine (VM) allocation strategies aiming to optimize the energy consumption of both cooling system and IT resources while meeting Service Level Agreements (SLAs). The default CloudSim architecture is extended by adding classes which contain mathematical models of the thermal processes within the server room. Furthermore, four new VM allocation policies that consider the cooling system energy consumption are developed based on the thermal and cooling system models. Finally, experiments are run to evaluate various metrics on a set of default CloudSim allocation algorithms and the proposed allocation algorithms. The results show that the proposed algorithms outperform the default CloudSim allocation strategy, Power Aware Best-Fit Decreasing (PABFD), in terms of overall energy consumption and the number of VM migrations, and have on average better results than other existing allocation strategies.","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115067742","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516650
V. Chifu, C. Pop, Adrian Birladeanu, Nicolae Dragoi, I. Salomie
This paper presents a Choice Function-based Constructive Hyper-Heuristic for generating personalized healthy menu recommendations based on a person’s nutrition, price and delivery time constraints. We model the problem of generating personalized healthy menus as an optimization problem for which the search space consists of a set of food packages, the solution is represented as a menu containing five food packages for each meal of the day, and the fitness function evaluates the degree to which a menu personalizes a person’s profile. In each step of the proposed hyper-heuristic’s iterative phase, a low level domain independent heuristic is chosen to be applied on the current menu, based on its affinity and competence. The hyper-heuristic has been evaluated on a set of persons’ profiles and a set of food packages developed in-house.
{"title":"Choice Function-based Constructive Hyper-Heuristic for Generating Personalized Healthy Menu Recommendations","authors":"V. Chifu, C. Pop, Adrian Birladeanu, Nicolae Dragoi, I. Salomie","doi":"10.1109/ICCP.2018.8516650","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516650","url":null,"abstract":"This paper presents a Choice Function-based Constructive Hyper-Heuristic for generating personalized healthy menu recommendations based on a person’s nutrition, price and delivery time constraints. We model the problem of generating personalized healthy menus as an optimization problem for which the search space consists of a set of food packages, the solution is represented as a menu containing five food packages for each meal of the day, and the fitness function evaluates the degree to which a menu personalizes a person’s profile. In each step of the proposed hyper-heuristic’s iterative phase, a low level domain independent heuristic is chosen to be applied on the current menu, based on its affinity and competence. The hyper-heuristic has been evaluated on a set of persons’ profiles and a set of food packages developed in-house.","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129269648","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516630
R. Brehar, F. Vancea, T. Mariţa, S. Nedevschi
Semantic segmentation in the context of traffic scenes has been vastly explored using different architectures for deep convolutional networks and color images. In the case of infrared images there is place for improvement and scientific contributions mainly due to the lack of data sets that contain baseline segmentations in the infrared domain. This paper proposes a method for real time infrared pedestrian segmentation using ERFNet. Within the context of the proposed method we study the effect of different basic image enhancement techniques on the performance of the segmentation. We enhance an existing dataset of infrared images with ground truth segmentations for pedestrians. Our experiments show that the proposed method is accurate and appropriate for real time applications.
{"title":"A Deep Learning Approach For Pedestrian Segmentation In Infrared Images","authors":"R. Brehar, F. Vancea, T. Mariţa, S. Nedevschi","doi":"10.1109/ICCP.2018.8516630","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516630","url":null,"abstract":"Semantic segmentation in the context of traffic scenes has been vastly explored using different architectures for deep convolutional networks and color images. In the case of infrared images there is place for improvement and scientific contributions mainly due to the lack of data sets that contain baseline segmentations in the infrared domain. This paper proposes a method for real time infrared pedestrian segmentation using ERFNet. Within the context of the proposed method we study the effect of different basic image enhancement techniques on the performance of the segmentation. We enhance an existing dataset of infrared images with ground truth segmentations for pedestrians. Our experiments show that the proposed method is accurate and appropriate for real time applications.","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"481 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113967169","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516645
C. Rusu, G. Czibula
Convolutional neural networks are effective supervised learning models which are widely used nowadays in various applications ranging from computer vision tasks such as image detection and classification, image captioning, to video classification. Even if the convolutional models are highly performant, a major drawback is given by their computationally expensiveness from the viewpoint of the required memory, additions and multiplications operations and thus are hardly portable on limited-resource devices. The purpose of this paper is to demonstrate the applicability of convolutional neural networks for low resource devices and to study their performance in real life scenarios. In this respect, with the major goal of preserving the performance, we propose a convolutional neural network model, called SimpLeNet, using distillation for image tagging that can run on low-resource devices such as smartphones, smartwatches, tablets or TVs. Experiments performed on MNIST data set for image classification emphasize the effectiveness of SimpLeNet, both in terms of model’s size reduction, as well as in terms of classification accuracy
{"title":"Optimizing Convolutional Neural Networks for low-resource devices","authors":"C. Rusu, G. Czibula","doi":"10.1109/ICCP.2018.8516645","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516645","url":null,"abstract":"Convolutional neural networks are effective supervised learning models which are widely used nowadays in various applications ranging from computer vision tasks such as image detection and classification, image captioning, to video classification. Even if the convolutional models are highly performant, a major drawback is given by their computationally expensiveness from the viewpoint of the required memory, additions and multiplications operations and thus are hardly portable on limited-resource devices. The purpose of this paper is to demonstrate the applicability of convolutional neural networks for low resource devices and to study their performance in real life scenarios. In this respect, with the major goal of preserving the performance, we propose a convolutional neural network model, called SimpLeNet, using distillation for image tagging that can run on low-resource devices such as smartphones, smartwatches, tablets or TVs. Experiments performed on MNIST data set for image classification emphasize the effectiveness of SimpLeNet, both in terms of model’s size reduction, as well as in terms of classification accuracy","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124509298","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516598
Emil Stefan Chifu, V. Chifu, C. Pop, A. Vlad, I. Salomie
This paper presents a technique for detecting the routine of the daily activities of a person and the deviations from this. The technique proposed has three main steps. The first step consists in identifying the daily living activities performed by a person by using two machine learning algorithms, one based on Decisions Trees and the other based on Random Forests. The second step consists in recognizing activity patterns corresponding to a daily routine by using the FP-Growth algorithm, while the third step computes the deviation from the daily activity routine of the person. The system proposed has been tested on the DaLiAc data set, which contains data collected from human subjects by using sensors based on accelerometers and gyroscopes.
{"title":"Machine Learning Based Technique for Detecting Daily Routine and Deviations","authors":"Emil Stefan Chifu, V. Chifu, C. Pop, A. Vlad, I. Salomie","doi":"10.1109/ICCP.2018.8516598","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516598","url":null,"abstract":"This paper presents a technique for detecting the routine of the daily activities of a person and the deviations from this. The technique proposed has three main steps. The first step consists in identifying the daily living activities performed by a person by using two machine learning algorithms, one based on Decisions Trees and the other based on Random Forests. The second step consists in recognizing activity patterns corresponding to a daily routine by using the FP-Growth algorithm, while the third step computes the deviation from the daily activity routine of the person. The system proposed has been tested on the DaLiAc data set, which contains data collected from human subjects by using sensors based on accelerometers and gyroscopes.","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"172 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124188917","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516580
Casian-Nicolae Marc, G. Czibula
Cytogenetics is a field of genetics investigating the relationships between the hereditary characteristics, structure and behavior of human chromosomes, as well as the medical and evolutionary repercussions of chromosomal abnormalities. Detecting the human karyotype and chromosomal anomalies could offer relevant information about human genetics and possible genetic disorders. This paper investigates an automatic solution for chromosomes classification and introduces an unsupervised learning approach KarySOM based on self-organizing maps for the problem of automatically human karyotyping, with the more general goal of uncovering chromosomal anomalies. The experimental evaluation of the proposed approach highlights its effectiveness for unsupervised classification of human chromosomes.
{"title":"KarySOM: An Unsupervised Learning based Approach for Human Karyotyping using Self-Organizing Maps","authors":"Casian-Nicolae Marc, G. Czibula","doi":"10.1109/ICCP.2018.8516580","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516580","url":null,"abstract":"Cytogenetics is a field of genetics investigating the relationships between the hereditary characteristics, structure and behavior of human chromosomes, as well as the medical and evolutionary repercussions of chromosomal abnormalities. Detecting the human karyotype and chromosomal anomalies could offer relevant information about human genetics and possible genetic disorders. This paper investigates an automatic solution for chromosomes classification and introduces an unsupervised learning approach KarySOM based on self-organizing maps for the problem of automatically human karyotyping, with the more general goal of uncovering chromosomal anomalies. The experimental evaluation of the proposed approach highlights its effectiveness for unsupervised classification of human chromosomes.","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116288496","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516617
Teodor Petrican, Andreea Valeria Vesa, Marcel Antal, Claudia Pop, T. Cioara, I. Anghel, I. Salomie
This paper tackles the problem of integrating household energy prosumers in Smart Energy Grids by analyzing a set of state-of-the-art energy forecasting techniques that allow individual or aggregated prosumers to evaluate their future energy demand and inform the Distributed System Operator (DSO) about potential grid imbalances. Thus, the DSO can perform a proactive strategy to manage the grid and avoid problems before they appear. The key element of this approach is the prediction technique, that must be accurate enough such that the resulting grid imbalances can be compensated in real-time. The paper evaluates a set of state-of-the-art statistical and Machine Learning (ML) prediction techniques, such as SARIMA, feed-forward and recurrent neural networks, support vector regression or ensemble prediction models, on real household historical energy demand logs by performing a feature selection process for each ML algorithm as to identify the best elements that influence the energy demand of a house. A set of experiments are performed on the REFIT Electrical Load Measurements data set evaluating each model’s performance with respect to the selected features. Among the evaluated algorithms, the Ensemble Prediction Model gives best prediction accuracy, showing a Mean Absolute Percentage Error (MAPE) of 14.4% followed by the SVM model with a MAPE of 15.4%.
{"title":"Evaluating Forecasting Techniques for Integrating Household Energy Prosumers into Smart Grids","authors":"Teodor Petrican, Andreea Valeria Vesa, Marcel Antal, Claudia Pop, T. Cioara, I. Anghel, I. Salomie","doi":"10.1109/ICCP.2018.8516617","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516617","url":null,"abstract":"This paper tackles the problem of integrating household energy prosumers in Smart Energy Grids by analyzing a set of state-of-the-art energy forecasting techniques that allow individual or aggregated prosumers to evaluate their future energy demand and inform the Distributed System Operator (DSO) about potential grid imbalances. Thus, the DSO can perform a proactive strategy to manage the grid and avoid problems before they appear. The key element of this approach is the prediction technique, that must be accurate enough such that the resulting grid imbalances can be compensated in real-time. The paper evaluates a set of state-of-the-art statistical and Machine Learning (ML) prediction techniques, such as SARIMA, feed-forward and recurrent neural networks, support vector regression or ensemble prediction models, on real household historical energy demand logs by performing a feature selection process for each ML algorithm as to identify the best elements that influence the energy demand of a house. A set of experiments are performed on the REFIT Electrical Load Measurements data set evaluating each model’s performance with respect to the selected features. Among the evaluated algorithms, the Ensemble Prediction Model gives best prediction accuracy, showing a Mean Absolute Percentage Error (MAPE) of 14.4% followed by the SVM model with a MAPE of 15.4%.","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132182980","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516591
Iuliana Gherghescu
The article contains a portion of the documentation commenced as part of a broader research project aimed at interactive intelligent spaces and interactive devices used in the design projects and in the development of spaces and stage items, costume or puppets. The attention is centered on the concept of interactivity as a component in shaping a new aesthetics in scenography. Interactivity, as a concept, will be defined in terms of performance design and related directly to the performer and how it will intervene in storytelling and change the spectator experience. The research focuses particularly on the design of spaces and sets for several types of performance including how classical spaces could be modified to become creative spaces by borrowing from architectural innovations and breakthroughs in intelligent surfaces and interactive facades
{"title":"Interactive spaces A change in scenographic aesthetics","authors":"Iuliana Gherghescu","doi":"10.1109/ICCP.2018.8516591","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516591","url":null,"abstract":"The article contains a portion of the documentation commenced as part of a broader research project aimed at interactive intelligent spaces and interactive devices used in the design projects and in the development of spaces and stage items, costume or puppets. The attention is centered on the concept of interactivity as a component in shaping a new aesthetics in scenography. Interactivity, as a concept, will be defined in terms of performance design and related directly to the performer and how it will intervene in storytelling and change the spectator experience. The research focuses particularly on the design of spaces and sets for several types of performance including how classical spaces could be modified to become creative spaces by borrowing from architectural innovations and breakthroughs in intelligent surfaces and interactive facades","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"107 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125564679","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516587
A. Alouache, Qing-he Wu
this paper investigates trajectory tracking for an autonomous nonholonomic wheeled mobile robot with virtual robot as reference trajectory. The standard proportional integral derivative (PID) is used for regulating the velocity of the follower robot such that the tracking errors are minimized between the follower and the reference trajectory. However using the PID controller solely for trajectory tracking produces poor results in the presence of noise or external disturbances. Hence genetic algorithms (GA) is applied in this paper to improve the performance of the PID controller in terms of control precision and speed of convergence. Moreover, communication between the follower and the virtual robot may fail very often in practice due to many raisons such as noise or external disturbances. Therefore, GA is applied again to predict the reference trajectory in case of communication disturbance. The simulation results demonstrate the effectiveness of the proposed GA-PID controller compared with the PID controller.
{"title":"Genetic Algorithms for Trajectory Tracking of Mobile Robot Based on PID Controller","authors":"A. Alouache, Qing-he Wu","doi":"10.1109/ICCP.2018.8516587","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516587","url":null,"abstract":"this paper investigates trajectory tracking for an autonomous nonholonomic wheeled mobile robot with virtual robot as reference trajectory. The standard proportional integral derivative (PID) is used for regulating the velocity of the follower robot such that the tracking errors are minimized between the follower and the reference trajectory. However using the PID controller solely for trajectory tracking produces poor results in the presence of noise or external disturbances. Hence genetic algorithms (GA) is applied in this paper to improve the performance of the PID controller in terms of control precision and speed of convergence. Moreover, communication between the follower and the virtual robot may fail very often in practice due to many raisons such as noise or external disturbances. Therefore, GA is applied again to predict the reference trajectory in case of communication disturbance. The simulation results demonstrate the effectiveness of the proposed GA-PID controller compared with the PID controller.","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"302 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121463153","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 : 2018-09-01DOI: 10.1109/ICCP.2018.8516603
C. Lungu, R. Potolea
Modern approaches that tackle super-resolution aim to train a generator network that transforms the low resolution image into a higher resolution one. The core learning capacity of these generator networks is given by stacks of well known image processing blocks such as VGG-16 [SZ14], ResNet[HZRS15] or Inception-v3 [SVI $^{+15]}$ blocks. In the light of recent advancements on the CIFAR-10 [KNH] benchmarks where DenseNet [HLW16] and later SparseNet [ZDD $^{+18]}$ proved superior performance over the architectures that used the formerly mentioned blocks, this paper aims to do a comparative study on the performance changes resulting when using DenseNet or SparseNet blocks in generator networks. We first replicate the results of [JAL16]. This work describes a generator network that uses a stack of four ResNet blocks. This stack is incorporated in two architectures for superresolution, one for x4 magnification and another one for x8. We then proceed and substitute them with DenseNet blocks and SparseNet blocks but keep the same overall training procedure. In order to ensure a fair comparison we adapt the number of blocks for each architecture in order to match the same amount of parameters on all architectures. In all cases the same optimization loss function is used, perceptual loss [JAL16], which for a given image yields a value that is a weighted sum of mean-squared-errors between filters of the target input and generated image evaluated on equivalent convolution layers of the last three blocks in the VGG-16 network (pretrained on the ImageNet [DDS $^{+09]}$ dataset). We monitor on all architectures the loss value, the number of epochs needed to reach the lowest loss, the artifacts generated by each network and the overall appearance of the reconstructions.
{"title":"Super-Resolution Generator Networks: A comparative study","authors":"C. Lungu, R. Potolea","doi":"10.1109/ICCP.2018.8516603","DOIUrl":"https://doi.org/10.1109/ICCP.2018.8516603","url":null,"abstract":"Modern approaches that tackle super-resolution aim to train a generator network that transforms the low resolution image into a higher resolution one. The core learning capacity of these generator networks is given by stacks of well known image processing blocks such as VGG-16 [SZ14], ResNet[HZRS15] or Inception-v3 [SVI $^{+15]}$ blocks. In the light of recent advancements on the CIFAR-10 [KNH] benchmarks where DenseNet [HLW16] and later SparseNet [ZDD $^{+18]}$ proved superior performance over the architectures that used the formerly mentioned blocks, this paper aims to do a comparative study on the performance changes resulting when using DenseNet or SparseNet blocks in generator networks. We first replicate the results of [JAL16]. This work describes a generator network that uses a stack of four ResNet blocks. This stack is incorporated in two architectures for superresolution, one for x4 magnification and another one for x8. We then proceed and substitute them with DenseNet blocks and SparseNet blocks but keep the same overall training procedure. In order to ensure a fair comparison we adapt the number of blocks for each architecture in order to match the same amount of parameters on all architectures. In all cases the same optimization loss function is used, perceptual loss [JAL16], which for a given image yields a value that is a weighted sum of mean-squared-errors between filters of the target input and generated image evaluated on equivalent convolution layers of the last three blocks in the VGG-16 network (pretrained on the ImageNet [DDS $^{+09]}$ dataset). We monitor on all architectures the loss value, the number of epochs needed to reach the lowest loss, the artifacts generated by each network and the overall appearance of the reconstructions.","PeriodicalId":259007,"journal":{"name":"2018 IEEE 14th International Conference on Intelligent Computer Communication and Processing (ICCP)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131772654","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：