Papers by Konstantinos Tzevanidis
Advances in Intelligent Systems and Computing, 2015
Markerless 3D tracking of hands in action or in interaction with objects provides rich informatio... more Markerless 3D tracking of hands in action or in interaction with objects provides rich information that can be used to interpret a number of human activities. In this paper, we review a number of relevant methods we have proposed. All of them focus on hands, objects and their interaction and follow a generative approach. The major strength of such an approach is the straightforward fashion in which arbitrarily complex priors can be easily incorporated towards solving the tracking problem and their capability to generalize to greater and/or different domains. The proposed generative approach is implemented in a single, unified computational framework. Keywords 3D hand tracking • 3D tracking of hand-object interactions • Modelbased 3D tracking
A Prototypical Interactive Exhibition for the Archaeological Museum of Thessaloniki
International Journal of Heritage in the Digital Era, 2013
In 2010, the Institute of Computer Science of the Foundation for Research and Technology-Hellas (... more In 2010, the Institute of Computer Science of the Foundation for Research and Technology-Hellas (ICS-FORTH) and the Archaeological Museum of Thessaloniki (AMTh) collaborated towards the creation of a special exhibition of prototypical interactive systems with subjects drawn from ancient Macedonia, named “Macedonia from fragments to pixels”. The exhibition comprises seven interactive systems based on the research outcomes of ICS-FORTH's Ambient Intelligence Programme. Up to the summer of 2012, more than 165.000 people have visited it. The paper initially provides some background information, including related previous research work, and then illustrates and discusses the development process that was followed for creating the exhibition. Subsequently, the technological and interactive characteristics of the project's outcomes (i.e., the interactive systems) are analysed and the complementary evaluation approaches followed are briefly described. Finally, some conclusions stemmi...
2013 International Symposium on Ubiquitous Virtual Reality, 2013
We present a method for articulated hand tracking that relies on visual input acquired by a calib... more We present a method for articulated hand tracking that relies on visual input acquired by a calibrated multicamera system. A state-of-the-art result on this problem has been presented in [12]. In that work, hand tracking is formulated as the minimization of an objective function that quantifies the discrepancy between a hand pose hypothesis and the observations. The objective function treats the observations from each camera view in an independent way. We follow the same general optimization framework but we choose to employ the visual hull [10] as the main observation cue, which results from the integration of information from all available views prior to optimization. We investigate the behavior of the resulting method in extensive experiments and in comparison with that of [12]. The obtained results demonstrate that for low levels of noise contamination, regardless of the number of cameras, the two methods perform comparably. The situation changes when noisy observations or as few as two cameras with short baselines are employed. In these cases, the proposed method is more accurate than that of [12]. Thus, the proposed method is preferable in real-world scenarios with noisy observations obtained from easy-to-deploy, stereo camera setups.
In 2010, the Institute of Computer Science of the Foundation for Research and Technology-Hellas (... more In 2010, the Institute of Computer Science of the Foundation for Research and Technology-Hellas (ICS-FORTH) and the Archaeological Museum of Thessaloniki (AMTh) collaborated towards the creation of a special exhibition of prototypical interactive systems with subjects drawn from ancient Macedonia, named "Macedonia from fragments to pixels". The exhibition comprises seven interactive systems based on the research outcomes of ICS-FORTH's Ambient Intelligence Programme. Up to the summer of 2012, more than 165.000 people have visited it. The paper initially provides some background information, including related previous research work, and then illustrates and discusses the development process that was followed for creating the exhibition. Subsequently, the technological and interactive characteristics of the project's outcomes (i.e., the interactive systems) are analysed and the complementary evaluation approaches followed are briefly described. Finally, some conclusi...
ECCV Workshops, 2010
Abstract. We present work on exploiting modern graphics hardware towards the real-time production... more Abstract. We present work on exploiting modern graphics hardware towards the real-time production of a textured 3D mesh representation of a scene observed by a multicamera system. The employed computational infrastructure consists of a network of four PC workstations each of which is connected to a pair of cameras. One of the PCs is equipped with a GPU that is used for parallel computations. The result of the processing is a list of texture mapped triangles representing the reconstructed surfaces. In contrast to previous ...
Transfer learning refers to the process of conveying experience from a simple task to another mor... more Transfer learning refers to the process of conveying experience from a simple task to another more complex (and related) task in order to reduce the amount of time that is required to learn the latter task. Typically, in a transfer learning procedure the agent learns a behavior in a source task, and it uses the gained knowledge in order to speed up the learning process in a target task. Reinforcement Learning algorithms are time expensive when they learn from scratch, especially in complex domains, and transfer learning comprises a suitable solution to speed up the training process. In this work we propose a method that decomposes the target task in several instances of the source task and uses them to extract an advised action for the target task. We evaluate the efficacy of the proposed approach in the robotic soccer Keepaway domain. The results demonstrate that the proposed method helps to reduce the training time of the target task.
This paper presents a computer vision system that supports non-instrumented, location-based inter... more This paper presents a computer vision system that supports non-instrumented, location-based interaction of multiple users with digital representations of large-scale artifacts. The proposed system is based on a camera network that observes multiple humans in front of a very large display. The acquired views are used to volumetrically reconstruct and track the humans robustly and in real time, even in crowded scenes and challenging human configurations. Given the frequent and accurate monitoring of humans in space and time, a dynamic and personalized textual/graphical annotation of the display can be achieved based on the location and the walk-through trajectory of each visitor. The proposed system has been successfully deployed in an archaeological museum, offering its visitors the capability to interact with and explore a digital representation of an ancient wall painting. This installation permits an extensive evaluation of the proposed system in terms of tracking robustness, comp...
This paper presents a computer vision system that supports non-instrumented, location-based inter... more This paper presents a computer vision system that supports non-instrumented, location-based interaction of multiple users with digital representations of large-scale artifacts. The proposed system is based on a camera network that observes multiple humans in front of a very large display. The acquired views are used to volumetrically reconstruct and track the humans robustly and in real time, even in crowded scenes and challenging human configurations. Given the frequent and accurate monitoring of humans in space and time, a dynamic and personalized textual/graphical annotation of the display can be achieved based on the location and the walk-through trajectory of each visitor. The proposed system has been successfully deployed in an archaeological museum, offering its visitors the capability to interact with and explore a digital representation of an ancient wall painting. This installation permits an extensive evaluation of the proposed system in terms of tracking robustness, comp...
Abstract. This paper describes the outcomes stemming from the work of a multidisciplinary R&D pro... more Abstract. This paper describes the outcomes stemming from the work of a multidisciplinary R&D project of ICS-FORTH, aiming to explore and experiment with novel interactive museum exhibits, and to assess their utility, usability and potential impact. More specifically, four interactive systems are presented in this paper which have been integrated, tested and evaluated in a dedicated, appropriately designed, laboratory space. The paper also discusses key issues stemming from experience and observations in the course of qualitative evaluation sessions with a large number of participants.
Transfer learning refers to the process of conveying experience from a simple task to another mor... more Transfer learning refers to the process of conveying experience from a simple task to another more complex (and related) task in order to reduce the amount of time that is required to learn the latter task. Typically, in a transfer learning procedure the agent learns a behavior in a source task, and it uses the gained knowledge in order to speed up the learning process in a target task. Reinforcement Learning algorithms are time expensive when they learn from scratch, especially in complex domains, and transfer learning comprises a suitable solution to speed up the training process. In this work we propose a method that decomposes the target task in several instances of the source task and uses them to extract an adviced action for the target task. We evaluate the efficacy of the proposed approach in the robotic soccer Keepaway domain. The results demonstrate that the proposed method helps to reduce the training time of the target task.
Tracking Hand Articulations: Relying on 3D Visual Hulls Versus Relying on Multiple 2D Cues
2013 International Symposium on Ubiquitous Virtual Reality, 2013
ABSTRACT We present a method for articulated hand tracking that relies on visual input acquired b... more ABSTRACT We present a method for articulated hand tracking that relies on visual input acquired by a calibrated multicamera system. A state-of-the-art result on this problem has been presented in [12]. In that work, hand tracking is formulated as the minimization of an objective function that quantifies the discrepancy between a hand pose hypothesis and the observations. The objective function treats the observations from each camera view in an independent way. We follow the same general optimization framework but we choose to employ the visual hull [10] as the main observation cue, which results from the integration of information from all available views prior to optimization. We investigate the behavior of the resulting method in extensive experiments and in comparison with that of [12]. The obtained results demonstrate that for low levels of noise contamination, regardless of the number of cameras, the two methods perform comparably. The situation changes when noisy observations or as few as two cameras with short baselines are employed. In these cases, the proposed method is more accurate than that of [12]. Thus, the proposed method is preferable in real-world scenarios with noisy oobservations obtained from easy-to-deploy, stereo camera setups.
Background modeling algorithms are commonly used in camera setups for foreground object detection... more Background modeling algorithms are commonly used in camera setups for foreground object detection. Typically, these algorithms need adjustment of their parameters towards achieving optimal performance in different scenarios and/or lighting conditions. This is a tedious process requiring considerable effort by expert users. In this work we propose a novel, fully automatic method for the tuning of foreground detection parameters in calibrated multicamera systems. The proposed method requires neither user intervention nor ground truth data. Given a set of such parameters, we define a fitness function based on the consensus built from the multicamera setup regarding whether points belong to the scene foreground or background. The maximization of this fitness function through Particle Swarm Optimization leads to the adjustment of the foreground detection parameters. Extensive experimental results confirm the effectiveness of the adopted approach.
Lecture Notes in Computer Science, 2012
The theme of this paper is an exhibition of prototypical interactive systems with subjects drawn ... more The theme of this paper is an exhibition of prototypical interactive systems with subjects drawn from ancient Macedonia, named "Macedonia from fragments to pixels". Since 2010, the exhibition is hosted by the Archaeological Museum of Thessaloniki and is open daily to the general public. Up to now, more than 165.000 people have visited it. The exhibition comprises 7 interactive systems which are based on some research outcomes of the Ambient Intelligence Programme of the Institute of Computer Science, Foundation for Research and Technology -Hellas. The digital content of these systems includes objects from the Museum's permanent collection and from Macedonia.
Lecture Notes in Computer Science, 2010
In this paper, the design and implementation of a hardware/software platform for parallel and dis... more In this paper, the design and implementation of a hardware/software platform for parallel and distributed multiview vision processing is presented. The platform is focused at supporting the monitoring of human presence in indoor environments. Its architecture is focused at increased throughput through process pipelining as well as at reducing communication costs and hardware requirements. Using this platform, we present efficient implementations of basic visual processes such as person tracking, textured visual hull computation and head pose estimation. Using the proposed platform multiview visual operations can be combined and third-party ones integrated, to ultimately facilitate the development of interactive applications that employ visual input. Computational performance is benchmarked comparatively to state of the art and the efficacy of the approach is qualitatively assessed in the context of already developed applications related to interactive environments.
Lecture Notes in Computer Science, 2011
This paper describes the outcomes stemming from the work of a multidisciplinary R&D project of IC... more This paper describes the outcomes stemming from the work of a multidisciplinary R&D project of ICS-FORTH, aiming to explore and experiment with novel interactive museum exhibits, and to assess their utility, usability and potential impact. More specifically, four interactive systems are presented in this paper which have been integrated, tested and evaluated in a dedicated, appropriately designed, laboratory space. The paper also discusses key issues stemming from experience and observations in the course of qualitative evaluation sessions with a large number of participants.
Machine Vision and Applications, 2013
This paper presents a computer vision system that supports non-instrumented, location-based inter... more This paper presents a computer vision system that supports non-instrumented, location-based interaction of multiple users with digital representations of large-scale artifacts. The proposed system is based on a camera network that observes multiple humans in front of a very large display. The acquired views are used to volumetrically reconstruct and track the humans robustly and in real time, even in crowded scenes and challenging human configurations. Given the frequent and accurate monitoring of humans in space and time, a dynamic and personalized textual/graphical annotation of the display can be achieved based on the location and the walk-through trajectory of each visitor. The proposed system has been successfully deployed in an archaeological museum, offering its visitors the capability to interact with and explore a digital representation of an ancient wall painting. This installation permits an extensive evaluation of the proposed system in terms of tracking robustness, computational performance and usability. Furthermore, it proves that computer vision technology can be effectively used to support non-instrumented interaction of humans with their environments in realistic settings.
Computer Vision and Image Understanding, 2011
Background modeling algorithms are commonly used in camera setups for foreground object detection... more Background modeling algorithms are commonly used in camera setups for foreground object detection. Typically, these algorithms need adjustment of their parameters towards achieving optimal performance in different scenarios and/or lighting conditions. This is a tedious process requiring considerable effort by expert users. In this work we propose a novel, fully automatic method for the tuning of foreground detection parameters in calibrated multicamera systems. The proposed method requires neither user intervention nor ground truth data. Given a set of such parameters, we define a fitness function based on the consensus built from the multicamera setup regarding whether points belong to the scene foreground or background. The maximization of this fitness function through Particle Swarm Optimization leads to the adjustment of the foreground detection parameters. Extensive experimental results confirm the effectiveness of the adopted approach.
Proceedings of The 8th …, May 10, 2009
Transfer learning refers to the process of conveying experience from a simple task to another mor... more Transfer learning refers to the process of conveying experience from a simple task to another more complex (and related) task in order to reduce the amount of time that is required to learn the latter task. Typically, in a transfer learning procedure the agent learns a behavior in a source task, and it uses the gained knowledge in order to speed up the learning process in a target task. Reinforcement Learning algorithms are time expensive when they learn from scratch, especially in complex domains, and transfer learning ...
Proc. of VAST, 2010
This paper presents a system that supports the exploration of digital representations of large-sc... more This paper presents a system that supports the exploration of digital representations of large-scale museum artifacts in through non-instrumented, location-based interaction. The system employs a state-of-the-art computer vision system, which localizes and tracks multiple visitors. The artifact is presented in a wall-sized projection screen and it is visually annotated with text and images according to the location as well as walkthrough trajectories of the tracked visitors. The system is evaluated in terms of computational performance, ...
A Generative Approach to Tracking Hands and Their Interaction with Objects
Advances in Intelligent Systems and Computing, 2015
Markerless 3D tracking of hands in action or in interaction with objects provides rich informatio... more Markerless 3D tracking of hands in action or in interaction with objects provides rich information that can be used to interpret a number of human activities. In this paper, we review a number of relevant methods we have proposed. All of them focus on hands, objects and their interaction and follow a generative approach. The major strength of such an approach is the straightforward fashion in which arbitrarily complex priors can be easily incorporated towards solving the tracking problem and their capability to generalize to greater and/or different domains. The proposed generative approach is implemented in a single, unified computational framework.
Uploads
Papers by Konstantinos Tzevanidis