Skip to main content

Klamer Schutte

Followers

5

Following

1

Public Views

Interests

Uploads

Papers by Klamer Schutte

Evaluation of super-resolution algorithms for mosaic hyperspectral imagery

Emerging Imaging and Sensing Technologies for Security and Defence IV

Semantic Reasoning in Zero Example Video Event Retrieval

ACM Transactions on Multimedia Computing, Communications, and Applications

Searching in digital video data for high-level events, such as a parade or a car accident, is cha... more Searching in digital video data for high-level events, such as a parade or a car accident, is challenging when the query is textual and lacks visual example images or videos. Current research in deep neural networks is highly beneficial for the retrieval of high-level events using visual examples, but without examples it is still hard to 1) determine which concepts are useful to pre-train (Vocabulary challenge); 2) which pre-trained concept detectors are relevant for a certain unseen high-level event (Concept Selection challenge). In our paper, we present our Semantic Event Retrieval System which 1) shows the importance of high-level concepts in a vocabulary for the retrieval of complex and generic high-level events and 2) uses a novel concept selection method (i-w2v) based on semantic embeddings. Our experiments on the international TRECVID Multimedia Event Detection benchmark show that a diverse vocabulary including high-level concepts improves performance on the retrieval of high-level events in videos and that our novel method outperforms a knowledge-based concept selection method.

Improving video event retrieval by user feedback

Multimedia Tools and Applications

In content based video retrieval videos are often indexed with semantic labels (concepts) using p... more In content based video retrieval videos are often indexed with semantic labels (concepts) using pre-trained classifiers. These pre-trained classifiers (concept detectors), are not perfect, and thus the labels are noisy. Additionally, the amount of pre-trained classifiers is limited. Often automatic methods cannot represent the query adequately in terms of the concepts available. This problem is also apparent in the retrieval of events, such as bike trick or birthday party. Our solution is to obtain user feedback. This user feedback can be provided on two levels: concept level and video level. We introduce the method Adaptive Relevance Feedback (ARF) on video level feedback. ARF is based on the classical Rocchio relevance feedback method from Information Retrieval. Furthermore, we explore Maaike de Boer

Probabilistic classification between foreground objects and background

Proceedings of the 17th International Conference on Pattern Recognition 2004 Icpr 2004, Aug 23, 2004

Tracking of deformable objects like humans is a basic operation in many surveillance applications... more Tracking of deformable objects like humans is a basic operation in many surveillance applications. Objects are detected as they enter the field of view of the camera and they are then tracked during the time they are visible. A problem with tracking deformable objects is that the shape of the object should be re-estimated for each frame. We propose a probabilistic framework combining object detection, tracking and shape deformation. We make use of the probabilities that a pixel belongs to the background, a new object or any of the known objects. Instead of using arbitrary thresholds for deciding to which class the pixel should be assigned we assign the pixel based on the Bayes criterion. Preliminary experiments show the classification error drops to about half the error of traditional approaches.

(EM) switch: a multi-hypothesis approach to (EM) background modelling

Ned Tijdschr Geneeskd, 2003

The detection of moving foreground objects is an important aspect of many vision applications, es... more The detection of moving foreground objects is an important aspect of many vision applications, especially those related to video-surveillance. The Expectation Maximisation (EM) algorithm is used in many of such applications to model the background and classify object pixels. In this paper two improvements to the EM algorithm will be given for object detection. First, it will show how to calculate background probabilities instead of binary foregroundbackground classifications. These probabilities will be compared to the foreground probabilities. Second, a multi-hypothesis approach will be introduced to decide when to update the model of the background (EMswitch). This way, the model of the background is not disturbed by passing objects. At the same time it gives a accurate description of the background statistics using one or more Gaussian kernels. The standard deviation is estimated more accurately then it would be estimated using an algorithm which only updates the background model for pixels which are classified as depicting background.

Hold Your Fire!: Preventing Fratricide in the Dismounted Soldier Domain

Since WWI, an estimated 15% of all soldiers killed in combat are attributed to fratricide, and re... more Since WWI, an estimated 15% of all soldiers killed in combat are attributed to fratricide, and recent military operations show no decline. A substantial amount concerns fratricide incidents between dismounted soldiers. However, most techniques introduced to prevent fratricide focus on inter-vehicle identification (IFF systems). In this feasibility study, we propose a decision support system that warns a dismounted soldier about to engage when there is a high risk of fratricide. The project has won the 2007 Innovation Game organized by the Dutch Ministry of Defense (MoD). The system will run on a new platform that is currently in development within the Dutch Soldier Modernisation Programme. This platform enables information exchange between soldiers and the Battlefield Management System (BMS), among which up-to-date soldier position information from GPS. Together with terrain information, also available in BMS, these soldier positions are taken into account when deriving an instant risk estimation for fratricide in the current shooting direction. The system is demonstrated using a simulation in which the village of Marnehuizen, built to train Military Operations on Urban Terrain (MOUT), serves as an example. The Dutch army has expressed great interest in the outcome of the study, and is currently investigating possibilities for actual implementation.

CCD Color Camera Characterization for Image Measurements

Ieee Transactions on Instrumentation and Measurement February 1 56 199 203, Feb 1, 2007

In this article, we will analyze a range of different types of cameras for its use in measurement... more In this article, we will analyze a range of different types of cameras for its use in measurements. We verify a general model of a charged coupled device camera using experiments. This model includes gain and offset, additive and multiplicative noise, and gamma correction. It is shown that for several cameras, except a typical consumer webcam, the general model holds. The associated model parameters are estimated. It is shown that for most cameras the model can be simplified under normal operating conditions by neglecting the dark current. We further show that the amount of additive noise is exceeded by the amount of multiplicative noise at intensity values larger than 10%-30% of the intensity range.

Correlations between 48 human actions improve their detection

Proceedings of the 21st International Conference on Pattern Recognition, 2012

Likelihood-based object detection and object tracking using color histograms and EM

Proceedings Icip International Conference on Image Processing, Aug 11, 2002

The topic of this paper is the integration of Expectation Maximization (EM) background modeling a... more The topic of this paper is the integration of Expectation Maximization (EM) background modeling and template matching using color histograms as templates to improve person tracking for surveillance applications. The tracked objects are humans, which are not rigid bodies. As such shape deformations of the objects must be allowed. For each frame, the decision has to be made which pixels belong to an object, and which do not. The integration of detection and tracking is done using a likelihood-based framework. This way the classification of pixels between background and object can be based on comparing likelihoods rather then separate thresholds. A demonstration of the proposed algorithm will be given.

Likelihood-based object tracking using color histograms and EM

Infsof, 2002

In?uence of signal-to-noise ratio and point spread function on limits of superresolution

Ipas, 2005

ABSTRACT This paper presents a method to predict the limit of possible resolution enhancement,giv... more ABSTRACT This paper presents a method to predict the limit of possible resolution enhancement,given a sequence of lowresolution images. Three important parameters influence the outcome of this limit: the total Point Spread Function (PSF), the Signal-to-Noise Ratio (SNR) and the number of input images. Although a large number of input images captured by a system with a narrow PSF and a high SNR are desirable, these conditions are often not achievable simultaneously. To improve the SNR, cameras are designed with near optimal quantum efficiency and maximum fill-factor. However, the latter widens the system PSF, which puts more weight on the deblurring part of a super-resolution (SR) reconstruction algorithm. This paper analyzes the contribution of each input parameters to the SR reconstruction and predicts the best attainable SR factor for given a camera setting. The predicted SR factor agrees well with an edge sharpness measure computed from the reconstructed SR images. A sufficient number of randomly positioned input images to achieve this limit for a given scene can also be derived assuming Gaussian noise and registration errors. Keywords: super-resolution limit, fill-factor, under-sampling, deconvolution limit, registration error.

An Edge Labeling Approach to Concave Polygon Clipping

Tog, Sep 3, 1995

EMswitch: A Multi-Hypothesis Approach to EM Background

The detection of moving foreground objects is an important aspect of many vision applications, es... more The detection of moving foreground objects is an important aspect of many vision applications, especially those related to video-surveillance. The Expectation Maximisation (EM) algorithm is used in many of such applications to model the background and classify object pixels. In this paper two improvements to the EM algorithm will be given for object detection. First, it will show how to calculate background probabilities instead of binary foregroundbackground classifications. These probabilities will be compared to the foreground probabilities. Second, a multi-hypothesis approach will be introduced to decide when to update the model of the background (EMswitch). This way, the model of the background is not disturbed by passing objects. At the same time it gives a accurate description of the background statistics using one or more Gaussian kernels. The standard deviation is estimated more accurately then it would be estimated using an algorithm which only updates the background model for pixels which are classified as depicting background.

Image Processing : An Enabler for Future EO System Concepts

4Th International Symposium on Optronics in Defense and Security Optro 2010 3 5 February Paris France, 2010

The current state of the art in electro-optics provides systems with high image quality for assoc... more The current state of the art in electro-optics provides systems with high image quality for associated prices and less expensive systems with subsequent lower performance. This keynote will expound how image processing enables to obtain high quality imagery while utilizing affordable system components. Furthermore, where conventional high-end EO systems are hindered by too much clutter and badly require often unavailable operator assistance, it will be shown that image processing can autonomously generate that information as needed in today's missions.

Method of improving the resolution of a moving object in a digital image sequence

Method of improving the resolution of a small moving object in a digital image sequence comprises... more Method of improving the resolution of a small moving object in a digital image sequence comprises the steps of: constructing (101) a high-resolution image background model, detecting (102) the moving object using the high-resolution image background model, fitting (103) a model-based trajectory for object registration, and producing (104) a high-resolution object description. The step of producing a high-resolution object description involves an iterative optimisation of a cost function (109) based upon a polygonal model of an edge of the moving object. The cost function is preferably also based upon a high resolution intensity description. The iterative optimisation of the cost function may involve a polygon description parameter and/or an intensity parameter

Image enhancement technology research for army applications

Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XXIV, 2013

Recognition and identification ranges are limited to the quality of the images. Both the received... more Recognition and identification ranges are limited to the quality of the images. Both the received contrast and the spatial resolution determine if objects are recognizable. Several aspects affect the image quality. First of all the sensor itself. The image quality depends on the size of the infrared detector array and the sensitivity. Second, also the intervening atmosphere, in particular over longer ranges, has an impact on the image quality. It degrades the contrast, due to transmission effects, as well as it influences the resolution, due to turbulence blur, of the image. We present studies in the field of infrared image enhancement. Several techniques are described: noise reduction, super resolution, turbulence compensation, contrast enhancement, stabilization. These techniques operate in real-time on COTS/MOTS platforms. They are especially effective in the army theatre, where long horizontal paths, and short line-of-sight limited urban operations are both present. Application of these techniques on observation masts, such as on military camp sites, and on UAVs and moving ground vehicles are discussed. Examples will be presented from several trials in which these techniques were demonstrated, including the presentation of test results.

Performance of optimal registration 12 EURASIP Journal on Applied Signal Processing estimators

Infrared polarisation measurements and modelling aplied to surface laid anti-personell landmines

Optical Engineering

Linear polarization of thermal infrared (TIR) radiation occurs when radiation is reflected or emi... more Linear polarization of thermal infrared (TIR) radiation occurs when radiation is reflected or emitted from a smooth surface (such as the top of a landmine) and observed from a grazing angle. The background (soil and vegetation) is generally much rougher and therefore shows less pronounced linearly polarized radiation. This difference in polarization is utilized to enhance detection of landmines using TIR cameras. A setup has been constructed for the acquisition of polarized TIR images. This setup contains a polarization filter that rotates synchronously to the frame sync of the camera. Either a long wave infrared (LWIR) or a mid wave infrared (MWIR) camera can be mounted behind the rotating polarization filter. The synchronization allows a sequence of images to be taken with a predefined constant angle of rotation between the images. Using this image sequence, three independent Stokes images are calculated, consisting of the unpolarized radiance, the difference between vertically an...

The TOSO data is the (toy and office-supplies objects) dataset. It contains a set of 145 test ima... more The TOSO data is the (toy and office-supplies objects) dataset. It contains a set of 145 test images used for the GOOSE demonstrator as described in: "Interactive detection of incrementally learned concepts in images with ranking and semantic query interpretation", Klamer Schutte , Henri Bouma , John Schavemaker , Laura Daniele , Maya Sappelli, Gijs Koot , Pieter Eendebak , George Azzopardi , Martijn Spitters , Maaike de Boer , Maarten Kruithof , Paul Brandt To be published 13th International Workshop on Content-Based Multimedia Indexing (CBMI) June 10.-12. 2015, Prague, Czech Republic

Evaluation of super-resolution algorithms for mosaic hyperspectral imagery

Emerging Imaging and Sensing Technologies for Security and Defence IV

Semantic Reasoning in Zero Example Video Event Retrieval

ACM Transactions on Multimedia Computing, Communications, and Applications

Searching in digital video data for high-level events, such as a parade or a car accident, is cha... more Searching in digital video data for high-level events, such as a parade or a car accident, is challenging when the query is textual and lacks visual example images or videos. Current research in deep neural networks is highly beneficial for the retrieval of high-level events using visual examples, but without examples it is still hard to 1) determine which concepts are useful to pre-train (Vocabulary challenge); 2) which pre-trained concept detectors are relevant for a certain unseen high-level event (Concept Selection challenge). In our paper, we present our Semantic Event Retrieval System which 1) shows the importance of high-level concepts in a vocabulary for the retrieval of complex and generic high-level events and 2) uses a novel concept selection method (i-w2v) based on semantic embeddings. Our experiments on the international TRECVID Multimedia Event Detection benchmark show that a diverse vocabulary including high-level concepts improves performance on the retrieval of high-level events in videos and that our novel method outperforms a knowledge-based concept selection method.

Improving video event retrieval by user feedback

Multimedia Tools and Applications

In content based video retrieval videos are often indexed with semantic labels (concepts) using p... more In content based video retrieval videos are often indexed with semantic labels (concepts) using pre-trained classifiers. These pre-trained classifiers (concept detectors), are not perfect, and thus the labels are noisy. Additionally, the amount of pre-trained classifiers is limited. Often automatic methods cannot represent the query adequately in terms of the concepts available. This problem is also apparent in the retrieval of events, such as bike trick or birthday party. Our solution is to obtain user feedback. This user feedback can be provided on two levels: concept level and video level. We introduce the method Adaptive Relevance Feedback (ARF) on video level feedback. ARF is based on the classical Rocchio relevance feedback method from Information Retrieval. Furthermore, we explore Maaike de Boer

Probabilistic classification between foreground objects and background

Proceedings of the 17th International Conference on Pattern Recognition 2004 Icpr 2004, Aug 23, 2004

Tracking of deformable objects like humans is a basic operation in many surveillance applications... more Tracking of deformable objects like humans is a basic operation in many surveillance applications. Objects are detected as they enter the field of view of the camera and they are then tracked during the time they are visible. A problem with tracking deformable objects is that the shape of the object should be re-estimated for each frame. We propose a probabilistic framework combining object detection, tracking and shape deformation. We make use of the probabilities that a pixel belongs to the background, a new object or any of the known objects. Instead of using arbitrary thresholds for deciding to which class the pixel should be assigned we assign the pixel based on the Bayes criterion. Preliminary experiments show the classification error drops to about half the error of traditional approaches.

(EM) switch: a multi-hypothesis approach to (EM) background modelling

Ned Tijdschr Geneeskd, 2003

The detection of moving foreground objects is an important aspect of many vision applications, es... more The detection of moving foreground objects is an important aspect of many vision applications, especially those related to video-surveillance. The Expectation Maximisation (EM) algorithm is used in many of such applications to model the background and classify object pixels. In this paper two improvements to the EM algorithm will be given for object detection. First, it will show how to calculate background probabilities instead of binary foregroundbackground classifications. These probabilities will be compared to the foreground probabilities. Second, a multi-hypothesis approach will be introduced to decide when to update the model of the background (EMswitch). This way, the model of the background is not disturbed by passing objects. At the same time it gives a accurate description of the background statistics using one or more Gaussian kernels. The standard deviation is estimated more accurately then it would be estimated using an algorithm which only updates the background model for pixels which are classified as depicting background.

Hold Your Fire!: Preventing Fratricide in the Dismounted Soldier Domain

Since WWI, an estimated 15% of all soldiers killed in combat are attributed to fratricide, and re... more Since WWI, an estimated 15% of all soldiers killed in combat are attributed to fratricide, and recent military operations show no decline. A substantial amount concerns fratricide incidents between dismounted soldiers. However, most techniques introduced to prevent fratricide focus on inter-vehicle identification (IFF systems). In this feasibility study, we propose a decision support system that warns a dismounted soldier about to engage when there is a high risk of fratricide. The project has won the 2007 Innovation Game organized by the Dutch Ministry of Defense (MoD). The system will run on a new platform that is currently in development within the Dutch Soldier Modernisation Programme. This platform enables information exchange between soldiers and the Battlefield Management System (BMS), among which up-to-date soldier position information from GPS. Together with terrain information, also available in BMS, these soldier positions are taken into account when deriving an instant risk estimation for fratricide in the current shooting direction. The system is demonstrated using a simulation in which the village of Marnehuizen, built to train Military Operations on Urban Terrain (MOUT), serves as an example. The Dutch army has expressed great interest in the outcome of the study, and is currently investigating possibilities for actual implementation.

CCD Color Camera Characterization for Image Measurements

Ieee Transactions on Instrumentation and Measurement February 1 56 199 203, Feb 1, 2007

In this article, we will analyze a range of different types of cameras for its use in measurement... more In this article, we will analyze a range of different types of cameras for its use in measurements. We verify a general model of a charged coupled device camera using experiments. This model includes gain and offset, additive and multiplicative noise, and gamma correction. It is shown that for several cameras, except a typical consumer webcam, the general model holds. The associated model parameters are estimated. It is shown that for most cameras the model can be simplified under normal operating conditions by neglecting the dark current. We further show that the amount of additive noise is exceeded by the amount of multiplicative noise at intensity values larger than 10%-30% of the intensity range.

Correlations between 48 human actions improve their detection

Proceedings of the 21st International Conference on Pattern Recognition, 2012

Likelihood-based object detection and object tracking using color histograms and EM

Proceedings Icip International Conference on Image Processing, Aug 11, 2002

The topic of this paper is the integration of Expectation Maximization (EM) background modeling a... more The topic of this paper is the integration of Expectation Maximization (EM) background modeling and template matching using color histograms as templates to improve person tracking for surveillance applications. The tracked objects are humans, which are not rigid bodies. As such shape deformations of the objects must be allowed. For each frame, the decision has to be made which pixels belong to an object, and which do not. The integration of detection and tracking is done using a likelihood-based framework. This way the classification of pixels between background and object can be based on comparing likelihoods rather then separate thresholds. A demonstration of the proposed algorithm will be given.

Likelihood-based object tracking using color histograms and EM

Infsof, 2002

In?uence of signal-to-noise ratio and point spread function on limits of superresolution

Ipas, 2005

ABSTRACT This paper presents a method to predict the limit of possible resolution enhancement,giv... more ABSTRACT This paper presents a method to predict the limit of possible resolution enhancement,given a sequence of lowresolution images. Three important parameters influence the outcome of this limit: the total Point Spread Function (PSF), the Signal-to-Noise Ratio (SNR) and the number of input images. Although a large number of input images captured by a system with a narrow PSF and a high SNR are desirable, these conditions are often not achievable simultaneously. To improve the SNR, cameras are designed with near optimal quantum efficiency and maximum fill-factor. However, the latter widens the system PSF, which puts more weight on the deblurring part of a super-resolution (SR) reconstruction algorithm. This paper analyzes the contribution of each input parameters to the SR reconstruction and predicts the best attainable SR factor for given a camera setting. The predicted SR factor agrees well with an edge sharpness measure computed from the reconstructed SR images. A sufficient number of randomly positioned input images to achieve this limit for a given scene can also be derived assuming Gaussian noise and registration errors. Keywords: super-resolution limit, fill-factor, under-sampling, deconvolution limit, registration error.

An Edge Labeling Approach to Concave Polygon Clipping

Tog, Sep 3, 1995

EMswitch: A Multi-Hypothesis Approach to EM Background

The detection of moving foreground objects is an important aspect of many vision applications, es... more The detection of moving foreground objects is an important aspect of many vision applications, especially those related to video-surveillance. The Expectation Maximisation (EM) algorithm is used in many of such applications to model the background and classify object pixels. In this paper two improvements to the EM algorithm will be given for object detection. First, it will show how to calculate background probabilities instead of binary foregroundbackground classifications. These probabilities will be compared to the foreground probabilities. Second, a multi-hypothesis approach will be introduced to decide when to update the model of the background (EMswitch). This way, the model of the background is not disturbed by passing objects. At the same time it gives a accurate description of the background statistics using one or more Gaussian kernels. The standard deviation is estimated more accurately then it would be estimated using an algorithm which only updates the background model for pixels which are classified as depicting background.

Image Processing : An Enabler for Future EO System Concepts

4Th International Symposium on Optronics in Defense and Security Optro 2010 3 5 February Paris France, 2010

The current state of the art in electro-optics provides systems with high image quality for assoc... more The current state of the art in electro-optics provides systems with high image quality for associated prices and less expensive systems with subsequent lower performance. This keynote will expound how image processing enables to obtain high quality imagery while utilizing affordable system components. Furthermore, where conventional high-end EO systems are hindered by too much clutter and badly require often unavailable operator assistance, it will be shown that image processing can autonomously generate that information as needed in today's missions.

Method of improving the resolution of a moving object in a digital image sequence

Method of improving the resolution of a small moving object in a digital image sequence comprises... more Method of improving the resolution of a small moving object in a digital image sequence comprises the steps of: constructing (101) a high-resolution image background model, detecting (102) the moving object using the high-resolution image background model, fitting (103) a model-based trajectory for object registration, and producing (104) a high-resolution object description. The step of producing a high-resolution object description involves an iterative optimisation of a cost function (109) based upon a polygonal model of an edge of the moving object. The cost function is preferably also based upon a high resolution intensity description. The iterative optimisation of the cost function may involve a polygon description parameter and/or an intensity parameter

Image enhancement technology research for army applications

Infrared Imaging Systems: Design, Analysis, Modeling, and Testing XXIV, 2013

Recognition and identification ranges are limited to the quality of the images. Both the received... more Recognition and identification ranges are limited to the quality of the images. Both the received contrast and the spatial resolution determine if objects are recognizable. Several aspects affect the image quality. First of all the sensor itself. The image quality depends on the size of the infrared detector array and the sensitivity. Second, also the intervening atmosphere, in particular over longer ranges, has an impact on the image quality. It degrades the contrast, due to transmission effects, as well as it influences the resolution, due to turbulence blur, of the image. We present studies in the field of infrared image enhancement. Several techniques are described: noise reduction, super resolution, turbulence compensation, contrast enhancement, stabilization. These techniques operate in real-time on COTS/MOTS platforms. They are especially effective in the army theatre, where long horizontal paths, and short line-of-sight limited urban operations are both present. Application of these techniques on observation masts, such as on military camp sites, and on UAVs and moving ground vehicles are discussed. Examples will be presented from several trials in which these techniques were demonstrated, including the presentation of test results.

Performance of optimal registration 12 EURASIP Journal on Applied Signal Processing estimators

Infrared polarisation measurements and modelling aplied to surface laid anti-personell landmines

Optical Engineering

Linear polarization of thermal infrared (TIR) radiation occurs when radiation is reflected or emi... more Linear polarization of thermal infrared (TIR) radiation occurs when radiation is reflected or emitted from a smooth surface (such as the top of a landmine) and observed from a grazing angle. The background (soil and vegetation) is generally much rougher and therefore shows less pronounced linearly polarized radiation. This difference in polarization is utilized to enhance detection of landmines using TIR cameras. A setup has been constructed for the acquisition of polarized TIR images. This setup contains a polarization filter that rotates synchronously to the frame sync of the camera. Either a long wave infrared (LWIR) or a mid wave infrared (MWIR) camera can be mounted behind the rotating polarization filter. The synchronization allows a sequence of images to be taken with a predefined constant angle of rotation between the images. Using this image sequence, three independent Stokes images are calculated, consisting of the unpolarized radiance, the difference between vertically an...

The TOSO data is the (toy and office-supplies objects) dataset. It contains a set of 145 test ima... more The TOSO data is the (toy and office-supplies objects) dataset. It contains a set of 145 test images used for the GOOSE demonstrator as described in: "Interactive detection of incrementally learned concepts in images with ranking and semantic query interpretation", Klamer Schutte , Henri Bouma , John Schavemaker , Laura Daniele , Maya Sappelli, Gijs Koot , Pieter Eendebak , George Azzopardi , Martijn Spitters , Maaike de Boer , Maarten Kruithof , Paul Brandt To be published 13th International Workshop on Content-Based Multimedia Indexing (CBMI) June 10.-12. 2015, Prague, Czech Republic