Farsi language is one of the dominant languages in middle-east. A lot of work has been done on Fa... more Farsi language is one of the dominant languages in middle-east. A lot of work has been done on Farsi retrieval systems. Local context analysis is a query expansion method to improve retrieval performance. In this paper we have tried to tune LCA for Farsi language. We used Hamshahri collection and 60 queries to tune three parameters in LCA method which
2018 Innovations in Intelligent Systems and Applications (INISTA)
Human action recognition and analysis has given life to a wide variety of real-world applications... more Human action recognition and analysis has given life to a wide variety of real-world applications, ranging from surveillance and human-computer interaction to patient monitoring and rehabilitation. Most action recognition systems, especially smart-home or assistive living applications, depend on network infrastructures for easy data fusion and integration of different sensing modalities. However, despite the fact that action recognition methods have extensively been evaluated for their accuracy and there is a consensus on the ways to provide quality of service in various network infrastructures, there is poor coverage of the inherent challenges of performing human action in real world network-based applications. In this work, we attempt to document these challenges based on representative, state of the art techniques and venture to report on the open issues that need to be resolved by new techniques aiming to provide viable real world applications.
2017 9th International Conference on Virtual Worlds and Games for Serious Applications (VS-Games), 2017
Exertion games form a vastly expanding field, crossing over to machine learning and user studies,... more Exertion games form a vastly expanding field, crossing over to machine learning and user studies, with studies of qualitative traits of actions, such as the player's level of expertise. In this work, we show how simple shape descriptors based on variance features fare on such a demanding task. We formulate two variance-based features and experiment on a demanding sports related dataset, captured with a Kinect sensor, in an action-specific k-NN classification scheme. Results show that simple shape features can produce meaningful results on determining a player's experience level, further encouraging their incorporation in more intricate schemes and real-world applications.
2018 IEEE Conference on Computational Intelligence and Games (CIG), 2018
Modeling and accurately analyzing human activities plays an important role, considering the rise ... more Modeling and accurately analyzing human activities plays an important role, considering the rise of modern applications in human-computer interaction and, more recently, exertion games. Especially in serious exergames aimed at tutoring (e.g. sports) or rehabilitation and physiotherapy, the need for accurate detection of the human body and its motion is uncompromising. However, modern human skeleton tracking techniques suffer from a variety of issues, such as jittering and sensitivity to original conditions. In this study we show how a simple yet effective fairing pipeline on an inherently noisy dataset can produce data capable for precise experimentation with state-of-the-art human action modeling algorithms.
Although the amount of raw surgical videos, namely videos captured during surgical interventions,... more Although the amount of raw surgical videos, namely videos captured during surgical interventions, is growing fast, automatic retrieval and search remains a challenge. This is mainly due to the nature of the content, i.e. visually non-consistent tissue, diversity of internal organs, abrupt viewpoint changes and illumination variation. We propose a framework for retrieving surgical videos and a protocol for evaluating the results. The method is composed of temporal shot segmentation and representation based on deep features, and the protocol introduces novel criteria to the field. The experimental results prove the superiority of the proposed method and highlight the path towards a more effective protocol for evaluating surgical videos.
2017 8th International Conference on Information, Intelligence, Systems & Applications (IISA), 2017
Identification of unintentional falls is a critical application in smart assistive environments, ... more Identification of unintentional falls is a critical application in smart assistive environments, especially in the context of elderly care. However, visually discriminating between falls and fall-like, intentional activities is a challenging task. In this paper, we propose the utilization of a novel feature extraction scheme based on the newly formulated 3D Cylindrical Trace Transform, on spatio-temporal interest points, for the task of fall detection. Using this pipeline, we are able to produce features invariant to occlusion, viewpoint, camera placement and other distortions. Experimentation on two publicly available datasets, on a number of different conditions, proved the efficiency of the proposed methodology for the task at hand.
2016 Digital Media Industry & Academic Forum (DMIAF), 2016
A system for efficient multimedia content analysis and automatic annotation is presented in this ... more A system for efficient multimedia content analysis and automatic annotation is presented in this paper. The system is able to identify objects in videos and annotate them with metadata. It includes three modules: the first provides detection and recognition of faces; the second provides generic object detection, based on a deep convolutional neural network; the third provides automated location estimation and landmark recognition based on the state-of-the-art technologies of Bag-of-Words and RANSAC. The system has been developed and successfully tested in the framework of the EC Horizon 2020 Mecanex project, targeting advertising and campaign production markets.
Proceedings of the 9th Hellenic Conference on Artificial Intelligence, 2016
In this paper we present a system for the semantic representation, enrichment and querying of fil... more In this paper we present a system for the semantic representation, enrichment and querying of film content. The system combines state-of-the-art knowledge representation and reasoning techniques, adapted to film content, with visual analysis techniques for point of interest detection. The resulting system provides search services for specific objects and entities in film scenes and shots. The experimental results that we present illustrate the good performance of the proposed approach.
Due to the growing use of human action recognition in every day life applications, it has become ... more Due to the growing use of human action recognition in every day life applications, it has become one of the very hot topics in image analysis and pattern recognition. This paper presents a new feature extraction method for human action recognition. The method is based on the extraction of Trace transforms from binarized silhouettes, representing different stages of a single action period. A final history template com-posed from the above transforms, represents the whole se-quence containing much of the valuable spatio-temporal in-formation contained in a human action. The new method takes advantage of the natural specifications of the specific Trace transform, such as noise robustness, translation invari-ance and scalability easiness and produces effective, simple and fast created features. Classification experiments performed on KTH action database using Radial Basis Function (RBF) Kernel SVM, provided very competitive results indicating the potentials of the proposed technique.
Farsi language is one of the dominant languages in middle-east. A lot of work has been done on Fa... more Farsi language is one of the dominant languages in middle-east. A lot of work has been done on Farsi retrieval systems. Local context analysis is a query expansion method to improve retrieval performance. In this paper we have tried to tune LCA for Farsi language. We used Hamshahri collection and 60 queries to tune three parameters in LCA method which
2018 Innovations in Intelligent Systems and Applications (INISTA)
Human action recognition and analysis has given life to a wide variety of real-world applications... more Human action recognition and analysis has given life to a wide variety of real-world applications, ranging from surveillance and human-computer interaction to patient monitoring and rehabilitation. Most action recognition systems, especially smart-home or assistive living applications, depend on network infrastructures for easy data fusion and integration of different sensing modalities. However, despite the fact that action recognition methods have extensively been evaluated for their accuracy and there is a consensus on the ways to provide quality of service in various network infrastructures, there is poor coverage of the inherent challenges of performing human action in real world network-based applications. In this work, we attempt to document these challenges based on representative, state of the art techniques and venture to report on the open issues that need to be resolved by new techniques aiming to provide viable real world applications.
2017 9th International Conference on Virtual Worlds and Games for Serious Applications (VS-Games), 2017
Exertion games form a vastly expanding field, crossing over to machine learning and user studies,... more Exertion games form a vastly expanding field, crossing over to machine learning and user studies, with studies of qualitative traits of actions, such as the player's level of expertise. In this work, we show how simple shape descriptors based on variance features fare on such a demanding task. We formulate two variance-based features and experiment on a demanding sports related dataset, captured with a Kinect sensor, in an action-specific k-NN classification scheme. Results show that simple shape features can produce meaningful results on determining a player's experience level, further encouraging their incorporation in more intricate schemes and real-world applications.
2018 IEEE Conference on Computational Intelligence and Games (CIG), 2018
Modeling and accurately analyzing human activities plays an important role, considering the rise ... more Modeling and accurately analyzing human activities plays an important role, considering the rise of modern applications in human-computer interaction and, more recently, exertion games. Especially in serious exergames aimed at tutoring (e.g. sports) or rehabilitation and physiotherapy, the need for accurate detection of the human body and its motion is uncompromising. However, modern human skeleton tracking techniques suffer from a variety of issues, such as jittering and sensitivity to original conditions. In this study we show how a simple yet effective fairing pipeline on an inherently noisy dataset can produce data capable for precise experimentation with state-of-the-art human action modeling algorithms.
Although the amount of raw surgical videos, namely videos captured during surgical interventions,... more Although the amount of raw surgical videos, namely videos captured during surgical interventions, is growing fast, automatic retrieval and search remains a challenge. This is mainly due to the nature of the content, i.e. visually non-consistent tissue, diversity of internal organs, abrupt viewpoint changes and illumination variation. We propose a framework for retrieving surgical videos and a protocol for evaluating the results. The method is composed of temporal shot segmentation and representation based on deep features, and the protocol introduces novel criteria to the field. The experimental results prove the superiority of the proposed method and highlight the path towards a more effective protocol for evaluating surgical videos.
2017 8th International Conference on Information, Intelligence, Systems & Applications (IISA), 2017
Identification of unintentional falls is a critical application in smart assistive environments, ... more Identification of unintentional falls is a critical application in smart assistive environments, especially in the context of elderly care. However, visually discriminating between falls and fall-like, intentional activities is a challenging task. In this paper, we propose the utilization of a novel feature extraction scheme based on the newly formulated 3D Cylindrical Trace Transform, on spatio-temporal interest points, for the task of fall detection. Using this pipeline, we are able to produce features invariant to occlusion, viewpoint, camera placement and other distortions. Experimentation on two publicly available datasets, on a number of different conditions, proved the efficiency of the proposed methodology for the task at hand.
2016 Digital Media Industry & Academic Forum (DMIAF), 2016
A system for efficient multimedia content analysis and automatic annotation is presented in this ... more A system for efficient multimedia content analysis and automatic annotation is presented in this paper. The system is able to identify objects in videos and annotate them with metadata. It includes three modules: the first provides detection and recognition of faces; the second provides generic object detection, based on a deep convolutional neural network; the third provides automated location estimation and landmark recognition based on the state-of-the-art technologies of Bag-of-Words and RANSAC. The system has been developed and successfully tested in the framework of the EC Horizon 2020 Mecanex project, targeting advertising and campaign production markets.
Proceedings of the 9th Hellenic Conference on Artificial Intelligence, 2016
In this paper we present a system for the semantic representation, enrichment and querying of fil... more In this paper we present a system for the semantic representation, enrichment and querying of film content. The system combines state-of-the-art knowledge representation and reasoning techniques, adapted to film content, with visual analysis techniques for point of interest detection. The resulting system provides search services for specific objects and entities in film scenes and shots. The experimental results that we present illustrate the good performance of the proposed approach.
Due to the growing use of human action recognition in every day life applications, it has become ... more Due to the growing use of human action recognition in every day life applications, it has become one of the very hot topics in image analysis and pattern recognition. This paper presents a new feature extraction method for human action recognition. The method is based on the extraction of Trace transforms from binarized silhouettes, representing different stages of a single action period. A final history template com-posed from the above transforms, represents the whole se-quence containing much of the valuable spatio-temporal in-formation contained in a human action. The new method takes advantage of the natural specifications of the specific Trace transform, such as noise robustness, translation invari-ance and scalability easiness and produces effective, simple and fast created features. Classification experiments performed on KTH action database using Radial Basis Function (RBF) Kernel SVM, provided very competitive results indicating the potentials of the proposed technique.
Uploads
Papers by Stefanos Kollias