Pub Date : 2018-04-01DOI: 10.1109/SSIAI.2018.8470359
A. Raglin, Andre V Harrison, Douglas Summers-Stay
As agents (devices and software) are increasingly incorporated into every aspect of our lives, the research area of human-agent teaming has seen an increase in attention. This is particularly true considering the varied, dynamic, and fast pace operations Soldiers are currently facing and will be facing in the future. There is a common idea that, in the future, the speed of machines will far exceed a Soldiers’ ability to react or even comprehend the complex activities of their digital teammates, which is a concern. Uncertainty in this accelerated environment will present unique and unforeseen challenges that may potentially inhibit a Soldier’s ability to make decisions effectively and to efficiently decide fast enough to support the future battlefield optempo. To accelerate decision making in Army operations the military is relying on agents and enabling technologies such as complex systems that integrate intelligent sensor networks and autonomous devices. These systems-of- systems will be driven by machine learning enabled artificial intelligence algorithms and will form teams with human warfighters, where both must act as one unit to accomplish their mission. Explanations can provide key information about the data or behavior of complex systems to the human to aide human agent teaming.
{"title":"FUSED REASONING UNDER UNCERTAINTY FOR SOLDIER CENTRIC HUMAN-AGENT DECISION MAKING","authors":"A. Raglin, Andre V Harrison, Douglas Summers-Stay","doi":"10.1109/SSIAI.2018.8470359","DOIUrl":"https://doi.org/10.1109/SSIAI.2018.8470359","url":null,"abstract":"As agents (devices and software) are increasingly incorporated into every aspect of our lives, the research area of human-agent teaming has seen an increase in attention. This is particularly true considering the varied, dynamic, and fast pace operations Soldiers are currently facing and will be facing in the future. There is a common idea that, in the future, the speed of machines will far exceed a Soldiers’ ability to react or even comprehend the complex activities of their digital teammates, which is a concern. Uncertainty in this accelerated environment will present unique and unforeseen challenges that may potentially inhibit a Soldier’s ability to make decisions effectively and to efficiently decide fast enough to support the future battlefield optempo. To accelerate decision making in Army operations the military is relying on agents and enabling technologies such as complex systems that integrate intelligent sensor networks and autonomous devices. These systems-of- systems will be driven by machine learning enabled artificial intelligence algorithms and will form teams with human warfighters, where both must act as one unit to accomplish their mission. Explanations can provide key information about the data or behavior of complex systems to the human to aide human agent teaming.","PeriodicalId":422209,"journal":{"name":"2018 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131274709","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 : 2018-04-01DOI: 10.1109/SSIAI.2018.8470360
Maher Aldeghlawi, M. Velez-Reyes
This paper explores the use of column subset selection (CSS) for unsupervised band subset selection (BSS) in hyperspectral imaging. CSS is the problem of selecting the most independent columns of a matrix. Many deterministic and randomized algorithms have been proposed in the literature for CSS. This paper presents a comparison between different algorithms for CSS for BSS. The cosine of the angle between the range space spanned by the selected bands and the corresponding left singular vectors is used to evaluate the quality of the selected bands to represent the image. Numerical experiments are conducted using multispectral and hyperspectral data. Results show that SVDSS outperforms other deterministic algorithms while producing comparable results to a 2-stage randomized CSS in small images and in centered data. However, the randomized algorithm significantly outperforms deterministic approaches in large images.
{"title":"A Comparison of Column Subset Selection Methods for Unsupervised Band Subset Selection in Hyperspectral Imagery","authors":"Maher Aldeghlawi, M. Velez-Reyes","doi":"10.1109/SSIAI.2018.8470360","DOIUrl":"https://doi.org/10.1109/SSIAI.2018.8470360","url":null,"abstract":"This paper explores the use of column subset selection (CSS) for unsupervised band subset selection (BSS) in hyperspectral imaging. CSS is the problem of selecting the most independent columns of a matrix. Many deterministic and randomized algorithms have been proposed in the literature for CSS. This paper presents a comparison between different algorithms for CSS for BSS. The cosine of the angle between the range space spanned by the selected bands and the corresponding left singular vectors is used to evaluate the quality of the selected bands to represent the image. Numerical experiments are conducted using multispectral and hyperspectral data. Results show that SVDSS outperforms other deterministic algorithms while producing comparable results to a 2-stage randomized CSS in small images and in centered data. However, the randomized algorithm significantly outperforms deterministic approaches in large images.","PeriodicalId":422209,"journal":{"name":"2018 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI)","volume":"147 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132451888","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 : 2018-04-01DOI: 10.1109/SSIAI.2018.8470358
S. Fang, Chang Liu, Khalid Tahboub, F. Zhu, E. Delp, C. Boushey
Measuring accurate dietary intake, the process of determining what someone eats during the course of the day is considered to be an open research problem in the nutrition and health fields. We have developed image-based tools to automatically obtain accurate estimates of what foods and how much energy/nutrients a user consumes. In this work, we present a crowdsourcing tool we designed and implemented to collect large sets of relevant online food images. This tool can be used to locate food items and obtaining groundtruth segmentation masks associated with all the foods presented in an image. We present a systematic design for a crowdsourcing tool aiming specifically for the task of online food image collection and annotations with a detailed description. The crowdsoucing tool we designed is tailored to meet the needs of building a large image dataset for developing automatic dietary assessment tools in the nutrition and health fields.
{"title":"cTADA: The Design of a Crowdsourcing Tool for Online Food Image Identification and Segmentation","authors":"S. Fang, Chang Liu, Khalid Tahboub, F. Zhu, E. Delp, C. Boushey","doi":"10.1109/SSIAI.2018.8470358","DOIUrl":"https://doi.org/10.1109/SSIAI.2018.8470358","url":null,"abstract":"Measuring accurate dietary intake, the process of determining what someone eats during the course of the day is considered to be an open research problem in the nutrition and health fields. We have developed image-based tools to automatically obtain accurate estimates of what foods and how much energy/nutrients a user consumes. In this work, we present a crowdsourcing tool we designed and implemented to collect large sets of relevant online food images. This tool can be used to locate food items and obtaining groundtruth segmentation masks associated with all the foods presented in an image. We present a systematic design for a crowdsourcing tool aiming specifically for the task of online food image collection and annotations with a detailed description. The crowdsoucing tool we designed is tailored to meet the needs of building a large image dataset for developing automatic dietary assessment tools in the nutrition and health fields.","PeriodicalId":422209,"journal":{"name":"2018 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127830733","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 : 2018-04-01DOI: 10.1109/SSIAI.2018.8470346
H. Parmar, Xiangyu Liu, Hua Xie, B. Nutter, S. Mitra, L. R. Long, Sameer Kiran Antani
Functional Magnetic Resonance Imaging (fMRI) uses a noninvasive technique to study the functionality of the human brain by measuring the Blood Oxygenation Level Dependent (BOLD) signal and has been researched for decades. However, some potential problems still remain in achieving correct interpretation of BOLD-induced signals due to quite low signal levels, high noise levels, artifacts, lack of ground truth and a number of other inherent problems. We present here the development of a MATLAB based fMRI simulator (f-Sim) using digital phantom brain that generates quasi-realistic 4D fMRI volumes including modeled noise. Such 4D fMRI data can serve to hypothesize ground truth for experimentally acquired data under both task-evoked and resting state designs in investigation of localized or whole brain activation and functional connectivity patterns.
{"title":"f-Sim: A quasi-realistic fMRI simulation toolbox using digital brain phantom and modeled noise","authors":"H. Parmar, Xiangyu Liu, Hua Xie, B. Nutter, S. Mitra, L. R. Long, Sameer Kiran Antani","doi":"10.1109/SSIAI.2018.8470346","DOIUrl":"https://doi.org/10.1109/SSIAI.2018.8470346","url":null,"abstract":"Functional Magnetic Resonance Imaging (fMRI) uses a noninvasive technique to study the functionality of the human brain by measuring the Blood Oxygenation Level Dependent (BOLD) signal and has been researched for decades. However, some potential problems still remain in achieving correct interpretation of BOLD-induced signals due to quite low signal levels, high noise levels, artifacts, lack of ground truth and a number of other inherent problems. We present here the development of a MATLAB based fMRI simulator (f-Sim) using digital phantom brain that generates quasi-realistic 4D fMRI volumes including modeled noise. Such 4D fMRI data can serve to hypothesize ground truth for experimentally acquired data under both task-evoked and resting state designs in investigation of localized or whole brain activation and functional connectivity patterns.","PeriodicalId":422209,"journal":{"name":"2018 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131120615","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 : 2018-04-01DOI: 10.1109/SSIAI.2018.8470332
Tamal Batabyal
Automated detection of decentralized event dynamics together with the identification of irregular topology on which the event propagates is a challenging task, which has its application in areas such as geomorphology and video surveillance. The problem becomes severe when the underlying topology is time-varying and multiple events with varied scales exist on the same topology. Conventional research works separately to deal with the problems of detecting events and identifying topology. On one hand, the methodologies for event detection involving the graph-spectral response fail to perform spatiotemporal localization of events if the underlying topology is unknown. On the other hand, the algorithms which estimate the underlying graph topology assume only static nature of the events. In this work, we utilize vertex reinforcement based walks on the topology to simultaneously perform both the tasks by using a scalable and tractable algorithm. An ensemble of such walks recursively updates the event membership of each location in the topology followed by associating a spatial support of each event. Our approach shows improvement over state-of-the-art methods in terms of the spatiotemporal localization of decentralized events.
{"title":"DDT: Decentralized event Detection and Tracking using an ensemble of vertex-reinforced walks on a graph","authors":"Tamal Batabyal","doi":"10.1109/SSIAI.2018.8470332","DOIUrl":"https://doi.org/10.1109/SSIAI.2018.8470332","url":null,"abstract":"Automated detection of decentralized event dynamics together with the identification of irregular topology on which the event propagates is a challenging task, which has its application in areas such as geomorphology and video surveillance. The problem becomes severe when the underlying topology is time-varying and multiple events with varied scales exist on the same topology. Conventional research works separately to deal with the problems of detecting events and identifying topology. On one hand, the methodologies for event detection involving the graph-spectral response fail to perform spatiotemporal localization of events if the underlying topology is unknown. On the other hand, the algorithms which estimate the underlying graph topology assume only static nature of the events. In this work, we utilize vertex reinforcement based walks on the topology to simultaneously perform both the tasks by using a scalable and tractable algorithm. An ensemble of such walks recursively updates the event membership of each location in the topology followed by associating a spatial support of each event. Our approach shows improvement over state-of-the-art methods in terms of the spatiotemporal localization of decentralized events.","PeriodicalId":422209,"journal":{"name":"2018 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI)","volume":"41 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114131980","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 : 2018-04-01DOI: 10.1109/SSIAI.2018.8470351
Anush Ananthakumar
Face recognition systems are used in various fields such as biometric authentication, security enhancement, automobile control and user detection. This research is focused on developing a model to control a system using gestures, while simultaneously implementing continuous facial recognition to avoid unauthorized access. An effective face recognition system is developed and applied in conjunction with a gesture recognition system to control a wireless robot in real-time. The facial recognition system extracts the face using the Viola-Jones algorithm which utilizes Haar like features along with Adaboost training. This is followed by a Convolution Neural Network (CNN) based feature extractor and Support Vector Machine (SVM) to recognize the face. The gesture recognition is facilitated by using color segmentation, which involves extracting the skin tone of the detected face and using this to detect the position of hand. The gesture is obtained by tracking the hand using the Kanade-Lucas-Tomasi (KLT) algorithm. In this research, we additionally utilize a background subtraction model so as to extract the foreground and reduce the misclassifications. Such a technique highly improves the performance of the facial and gesture detector in complex and cluttered environments. The performance of the face detector was tested on different databases including the ORL, Caltech and Faces96 database. The efficacy of this system in controlling a robot in real-time has also been demonstrated in this research. It provides an accuracy of 94.44% for recognizing faces and greater than 90.8% for recognizing gestures in real-time applications. Such a system is seen to have superior performance coupled with a relatively lower computation requirement in comparison to existing techniques.
{"title":"Efficient Face And Gesture Recognition For Time Sensitive Application","authors":"Anush Ananthakumar","doi":"10.1109/SSIAI.2018.8470351","DOIUrl":"https://doi.org/10.1109/SSIAI.2018.8470351","url":null,"abstract":"Face recognition systems are used in various fields such as biometric authentication, security enhancement, automobile control and user detection. This research is focused on developing a model to control a system using gestures, while simultaneously implementing continuous facial recognition to avoid unauthorized access. An effective face recognition system is developed and applied in conjunction with a gesture recognition system to control a wireless robot in real-time. The facial recognition system extracts the face using the Viola-Jones algorithm which utilizes Haar like features along with Adaboost training. This is followed by a Convolution Neural Network (CNN) based feature extractor and Support Vector Machine (SVM) to recognize the face. The gesture recognition is facilitated by using color segmentation, which involves extracting the skin tone of the detected face and using this to detect the position of hand. The gesture is obtained by tracking the hand using the Kanade-Lucas-Tomasi (KLT) algorithm. In this research, we additionally utilize a background subtraction model so as to extract the foreground and reduce the misclassifications. Such a technique highly improves the performance of the facial and gesture detector in complex and cluttered environments. The performance of the face detector was tested on different databases including the ORL, Caltech and Faces96 database. The efficacy of this system in controlling a robot in real-time has also been demonstrated in this research. It provides an accuracy of 94.44% for recognizing faces and greater than 90.8% for recognizing gestures in real-time applications. Such a system is seen to have superior performance coupled with a relatively lower computation requirement in comparison to existing techniques.","PeriodicalId":422209,"journal":{"name":"2018 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126073610","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 : 2017-10-26DOI: 10.1109/SSIAI.2018.8470336
Y. Watkins, M. Sayeh, O. Iaroshenko, Garrett T. Kenyon
Bottleneck autoencoders have been actively researched as a solution to image compression tasks. However, we observed that bottleneck autoencoders produce subjectively low quality reconstructed images. In this work, we explore the ability of sparse coding to improve reconstructed image quality for the same degree of compression. We observe that sparse image compression produces visually superior reconstructed images and yields higher values of pixel-wise measures of reconstruction quality (PSNR and SSIM) compared to bottleneck autoencoders. In addition, we find that using alternative metrics that correlate better with human perception, such as feature perceptual loss and the classification accuracy, sparse image compression scores up to 18.06% and 2.7% higher, respectively, compared to bottleneck autoencoders. Although computationally much more intensive, we find that sparse coding is otherwise superior to bottleneck autoencoders for the same degree of compression.
{"title":"Image Compression: Sparse Coding vs. Bottleneck Autoencoders","authors":"Y. Watkins, M. Sayeh, O. Iaroshenko, Garrett T. Kenyon","doi":"10.1109/SSIAI.2018.8470336","DOIUrl":"https://doi.org/10.1109/SSIAI.2018.8470336","url":null,"abstract":"Bottleneck autoencoders have been actively researched as a solution to image compression tasks. However, we observed that bottleneck autoencoders produce subjectively low quality reconstructed images. In this work, we explore the ability of sparse coding to improve reconstructed image quality for the same degree of compression. We observe that sparse image compression produces visually superior reconstructed images and yields higher values of pixel-wise measures of reconstruction quality (PSNR and SSIM) compared to bottleneck autoencoders. In addition, we find that using alternative metrics that correlate better with human perception, such as feature perceptual loss and the classification accuracy, sparse image compression scores up to 18.06% and 2.7% higher, respectively, compared to bottleneck autoencoders. Although computationally much more intensive, we find that sparse coding is otherwise superior to bottleneck autoencoders for the same degree of compression.","PeriodicalId":422209,"journal":{"name":"2018 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI)","volume":"113 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114115426","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：