In this introduction to artificial neural networks we attempt to give an overview of the most imp... more In this introduction to artificial neural networks we attempt to give an overview of the most important types of neural networks employed in engineering and explain shortly how they operate and also how they relate to biological neural networks. The focus will mainly be on bio-inspired artificial neural network architectures and specifically to neo-perceptions. The latter belong to the family of convolutional neural networks. Their topology is somewhat similar to the one of the human visual cortex and they are based on receptive fields that allow, in combination with sub-sampling layers, for an improved robustness with regard to local spatial distortions. We demonstrate the application of artificial neural networks to face analysis--a domain we human beings are particularly good at, yet which poses great difficulties for digital computers running deterministic software programs.
In this introduction to artificial neural networks we attempt to give an overview of the most imp... more In this introduction to artificial neural networks we attempt to give an overview of the most important types of neural networks employed in engineering and explain shortly how they operate and also how they relate to biological neural networks. The focus will mainly be on bio-inspired artificial neural network architectures and specifically to neo-perceptions. The latter belong to the family of convolutional neural networks. Their topology is somewhat similar to the one of the human visual cortex and they are based on receptive fields that allow, in combination with sub-sampling layers, for an improved robustness with regard to local spatial distortions. We demonstrate the application of artificial neural networks to face analysis--a domain we human beings are particularly good at, yet which poses great difficulties for digital computers running deterministic software programs.
2012 IEEE Workshop on Mathematical Methods in Biomedical Image Analysis, 2012
This paper describes a method for tracking a tumour using the planar projections of fiducial mark... more This paper describes a method for tracking a tumour using the planar projections of fiducial markers as surrogates. The projections can originate from various sources such as a beam-eye view X-ray, a portal imager or a fluoroscope. The two-dimensional position of the fiducial markers in the planar image in conjunction with a population-based statistical motion model is used to accurately predict and track the motion of a target volume during treatment. The basic assumption is that the projected surrogate locations contain valuable information about the in-plane motion of the lesion whereas the statistical motion model helps to describe the unobserved out-of-plane motion of the target volume. We analysed the accuracy with regard to varying the camera position and uncertainty in the measurement of the surrogate positions to simulate image noise and camera registration errors. The experiments showed that the tumour motion can be robustly predicted with an accuracy of 2.6 mm over a wide range of target volumes and treatment field directions despite a measurement error of σ = 2 mm for the fiducials.
ABSTRACT Magnetic Resonance guided High Intensity Focused Ultrasound (MRgHIFU) is an emerging non... more ABSTRACT Magnetic Resonance guided High Intensity Focused Ultrasound (MRgHIFU) is an emerging non-invasive technology for the treatment of pathological tissue. The possibility of depositing sharply localised energy deep within the body without affecting the surrounding tissue requires the exact knowledge of the target’s position. The cyclic respiratory organ motion renders targeting challenging, as the treatment focus has to be continuously adapted according to the current target’s displacement in 3D space. In this paper, a combination of a patient-specific dynamic breath model and a population-based statistical motion model is used to compensate for the respiratory induced organ motion. The application of a population based statistical motion model replaces the acquisition of a patient-specific 3D motion model, nevertheless allowing for precise motion compensation.
Automatic face analysis has to cope with pose and light- ing variations. Especially pose variatio... more Automatic face analysis has to cope with pose and light- ing variations. Especially pose variations are difficult to tackle and many face analysis methods require the use of so- phisticated normalization procedures. We propose a data- driven face analysis approach that is not only capable of extracting features relevant to a given face analysis task, but is also robust with regard to face location changes and scale variations. This is achieved by deploying convolu- tional neural networks. We show that the use of multi- scale feature extractors and whole-field feature map sum- ming neurons allow to improve facial expression recogni- tion results, especially with test sets that feature scale, re- spectively, translation changes.
We discuss a neural networks-based face analysis approach that is able to cope with faces subject... more We discuss a neural networks-based face analysis approach that is able to cope with faces subject to pose and lighting variations. Especially head pose variations are difficult to tackle and many face analysis methods require the use of sophisticated normalization procedures. Data-driven shape and motion-based face analysis approaches are introduced that are not only capable of extracting features relevant to a given face analysis task, but are also robust with regard to translation and scale variations. This is achieved by deploying convolutional and time-delayed neural networks, which are either trained for face shape deformation or facial motion analysis.
Computerized human face processing (detection, recognition, synthesis) has known an intense resea... more Computerized human face processing (detection, recognition, synthesis) has known an intense research activity during the last few years. Applications involving human face recognition are very broad with an important commercial impacts. Human face ...
Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2012
In recent years, significant advances have been made towards compensating respiratory organ motio... more In recent years, significant advances have been made towards compensating respiratory organ motion for the treatment of tumours, e.g. for the liver. Among the most promising approaches are statistical population models of organ motion. In this paper we give an overview on our work in the field.We explain how 4D motion data can be acquired, how these motion models can then be built and applied in realistic scenarios. The application of the motion models is first shown on a case where 3D surrogate marker data is available. Then we will evaluate the prediction accuracy if only 2D and lastly 1D surrogate marker motion data is available. For all three scenarios we will give quantitative prediction accuracy results.
In this paper, we describe the prototype of an interactive museum guide. It runs on a tablet PC t... more In this paper, we describe the prototype of an interactive museum guide. It runs on a tablet PC that features a touchscreen, a webcam and a Bluetooth receiver. This guide recognises objects on display in museums based on images of the latter which are taken directly by the visitor. Furthermore, the computer can determine the visitor's lo- cation by receiving
We propose a new technique for a faster computation of the activities of the hidden layer units. ... more We propose a new technique for a faster computation of the activities of the hidden layer units. This has been demonstrated on face detection examples.
Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2012
In recent years, significant advances have been made towards compensating respiratory organ motio... more In recent years, significant advances have been made towards compensating respiratory organ motion for the treatment of tumours, e.g. for the liver. Among the most promising approaches are statistical population models of organ motion. In this paper we give an overview on our work in the field.We explain how 4D motion data can be acquired, how these motion models can then be built and applied in realistic scenarios. The application of the motion models is first shown on a case where 3D surrogate marker data is available. Then we will evaluate the prediction accuracy if only 2D and lastly 1D surrogate marker motion data is available. For all three scenarios we will give quantitative prediction accuracy results.
In this introduction to artificial neural networks we attempt to give an overview of the most imp... more In this introduction to artificial neural networks we attempt to give an overview of the most important types of neural networks employed in engineering and explain shortly how they operate and also how they relate to biological neural networks. The focus will mainly be on bio-inspired artificial neural network architectures and specifically to neo-perceptions. The latter belong to the family of convolutional neural networks. Their topology is somewhat similar to the one of the human visual cortex and they are based on receptive fields that allow, in combination with sub-sampling layers, for an improved robustness with regard to local spatial distortions. We demonstrate the application of artificial neural networks to face analysis--a domain we human beings are particularly good at, yet which poses great difficulties for digital computers running deterministic software programs.
In this introduction to artificial neural networks we attempt to give an overview of the most imp... more In this introduction to artificial neural networks we attempt to give an overview of the most important types of neural networks employed in engineering and explain shortly how they operate and also how they relate to biological neural networks. The focus will mainly be on bio-inspired artificial neural network architectures and specifically to neo-perceptions. The latter belong to the family of convolutional neural networks. Their topology is somewhat similar to the one of the human visual cortex and they are based on receptive fields that allow, in combination with sub-sampling layers, for an improved robustness with regard to local spatial distortions. We demonstrate the application of artificial neural networks to face analysis--a domain we human beings are particularly good at, yet which poses great difficulties for digital computers running deterministic software programs.
2012 IEEE Workshop on Mathematical Methods in Biomedical Image Analysis, 2012
This paper describes a method for tracking a tumour using the planar projections of fiducial mark... more This paper describes a method for tracking a tumour using the planar projections of fiducial markers as surrogates. The projections can originate from various sources such as a beam-eye view X-ray, a portal imager or a fluoroscope. The two-dimensional position of the fiducial markers in the planar image in conjunction with a population-based statistical motion model is used to accurately predict and track the motion of a target volume during treatment. The basic assumption is that the projected surrogate locations contain valuable information about the in-plane motion of the lesion whereas the statistical motion model helps to describe the unobserved out-of-plane motion of the target volume. We analysed the accuracy with regard to varying the camera position and uncertainty in the measurement of the surrogate positions to simulate image noise and camera registration errors. The experiments showed that the tumour motion can be robustly predicted with an accuracy of 2.6 mm over a wide range of target volumes and treatment field directions despite a measurement error of σ = 2 mm for the fiducials.
ABSTRACT Magnetic Resonance guided High Intensity Focused Ultrasound (MRgHIFU) is an emerging non... more ABSTRACT Magnetic Resonance guided High Intensity Focused Ultrasound (MRgHIFU) is an emerging non-invasive technology for the treatment of pathological tissue. The possibility of depositing sharply localised energy deep within the body without affecting the surrounding tissue requires the exact knowledge of the target’s position. The cyclic respiratory organ motion renders targeting challenging, as the treatment focus has to be continuously adapted according to the current target’s displacement in 3D space. In this paper, a combination of a patient-specific dynamic breath model and a population-based statistical motion model is used to compensate for the respiratory induced organ motion. The application of a population based statistical motion model replaces the acquisition of a patient-specific 3D motion model, nevertheless allowing for precise motion compensation.
Automatic face analysis has to cope with pose and light- ing variations. Especially pose variatio... more Automatic face analysis has to cope with pose and light- ing variations. Especially pose variations are difficult to tackle and many face analysis methods require the use of so- phisticated normalization procedures. We propose a data- driven face analysis approach that is not only capable of extracting features relevant to a given face analysis task, but is also robust with regard to face location changes and scale variations. This is achieved by deploying convolu- tional neural networks. We show that the use of multi- scale feature extractors and whole-field feature map sum- ming neurons allow to improve facial expression recogni- tion results, especially with test sets that feature scale, re- spectively, translation changes.
We discuss a neural networks-based face analysis approach that is able to cope with faces subject... more We discuss a neural networks-based face analysis approach that is able to cope with faces subject to pose and lighting variations. Especially head pose variations are difficult to tackle and many face analysis methods require the use of sophisticated normalization procedures. Data-driven shape and motion-based face analysis approaches are introduced that are not only capable of extracting features relevant to a given face analysis task, but are also robust with regard to translation and scale variations. This is achieved by deploying convolutional and time-delayed neural networks, which are either trained for face shape deformation or facial motion analysis.
Computerized human face processing (detection, recognition, synthesis) has known an intense resea... more Computerized human face processing (detection, recognition, synthesis) has known an intense research activity during the last few years. Applications involving human face recognition are very broad with an important commercial impacts. Human face ...
Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2012
In recent years, significant advances have been made towards compensating respiratory organ motio... more In recent years, significant advances have been made towards compensating respiratory organ motion for the treatment of tumours, e.g. for the liver. Among the most promising approaches are statistical population models of organ motion. In this paper we give an overview on our work in the field.We explain how 4D motion data can be acquired, how these motion models can then be built and applied in realistic scenarios. The application of the motion models is first shown on a case where 3D surrogate marker data is available. Then we will evaluate the prediction accuracy if only 2D and lastly 1D surrogate marker motion data is available. For all three scenarios we will give quantitative prediction accuracy results.
In this paper, we describe the prototype of an interactive museum guide. It runs on a tablet PC t... more In this paper, we describe the prototype of an interactive museum guide. It runs on a tablet PC that features a touchscreen, a webcam and a Bluetooth receiver. This guide recognises objects on display in museums based on images of the latter which are taken directly by the visitor. Furthermore, the computer can determine the visitor's lo- cation by receiving
We propose a new technique for a faster computation of the activities of the hidden layer units. ... more We propose a new technique for a faster computation of the activities of the hidden layer units. This has been demonstrated on face detection examples.
Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2012
In recent years, significant advances have been made towards compensating respiratory organ motio... more In recent years, significant advances have been made towards compensating respiratory organ motion for the treatment of tumours, e.g. for the liver. Among the most promising approaches are statistical population models of organ motion. In this paper we give an overview on our work in the field.We explain how 4D motion data can be acquired, how these motion models can then be built and applied in realistic scenarios. The application of the motion models is first shown on a case where 3D surrogate marker data is available. Then we will evaluate the prediction accuracy if only 2D and lastly 1D surrogate marker motion data is available. For all three scenarios we will give quantitative prediction accuracy results.
Uploads
Papers by Beat Fasel