Institute of Informatics
Topic
Deep neural networks for applications in image processing and computer vision
PhD. program
Informatics / Applied informatics / Robotics and Cybernetics
Name of the supervisor
Ing. Peter Malík, PhD.
Contact:
Receiving school
Faculty of Mathematics, Physics and Informatics, Comenius Uni in Bratislava
Annotation
Receiving schools (another possibilities):
- Faculty of Informatics and Information Technologies of STU in Bratislava (program: Applied Informatics)
- Faculty of Electrical Engineering and Information Technology of STU in Bratislava (program: Cybernetics)
- Faculty of Electrical Engineering and Informatics, Technical Uni of Košice (FEI TUKE, program: Informatics)
Artificial neural networks are becoming very widely used in practical applications. They are used to process a large amount of information contained in image, audio or text data. Their practical applications have varying levels of complexity, from the search of the most important data, complete data analysis, prediction and forecasting to the calculation of control signals in the form of an autonomous control system. Image processing is a specific area in which artificial neural networks achieve excellent results due to their ability to learn to recognize the most important features and characteristics of an image from a large number of image pixels. In simpler computer vision tasks, such as object classification, artificial neural networks perform better than humans. This has been proven in a number of application domains, including general object recognition, biometric data classification (face, gait), medical data recognition (X-ray, CT, MRI).
The current challenge for artificial neural network research is the more complex tasks of computer vision such as detection and instance segmentation in which human capabilities have not been surpassed. Equally important research task is to reduce hardware requirements of artificial neural network computation. Research into new efficient artificial neural network architectures has made a significant contribution in this area. It is the architecture of neural networks that still offers a wide scope for improvement as solutions are sought at a higher level of problem abstraction. Research thesis will focus on these research areas in order to develop new methods, algorithms or architectures that improve the parameters and capabilities of artificial neural networks and enable them to be effectively applied in practice. Priority research areas will be adapted after consultation and student participation in international competitions and active participation in international conferences are foreseen.
Key words: deep learning, convolution neural networks, neural networks architecture, detection, instance segmentation, image processing, computer vision, neural network inference
more info
year 2022/2023
- Faculty of Informatics and Information Technologies of STU in Bratislava (program: Applied Informatics)
- Faculty of Electrical Engineering and Information Technology of STU in Bratislava (program: Cybernetics)
- Faculty of Electrical Engineering and Informatics, Technical Uni of Košice (FEI TUKE, program: Informatics)
Artificial neural networks are becoming very widely used in practical applications. They are used to process a large amount of information contained in image, audio or text data. Their practical applications have varying levels of complexity, from the search of the most important data, complete data analysis, prediction and forecasting to the calculation of control signals in the form of an autonomous control system. Image processing is a specific area in which artificial neural networks achieve excellent results due to their ability to learn to recognize the most important features and characteristics of an image from a large number of image pixels. In simpler computer vision tasks, such as object classification, artificial neural networks perform better than humans. This has been proven in a number of application domains, including general object recognition, biometric data classification (face, gait), medical data recognition (X-ray, CT, MRI).
The current challenge for artificial neural network research is the more complex tasks of computer vision such as detection and instance segmentation in which human capabilities have not been surpassed. Equally important research task is to reduce hardware requirements of artificial neural network computation. Research into new efficient artificial neural network architectures has made a significant contribution in this area. It is the architecture of neural networks that still offers a wide scope for improvement as solutions are sought at a higher level of problem abstraction. Research thesis will focus on these research areas in order to develop new methods, algorithms or architectures that improve the parameters and capabilities of artificial neural networks and enable them to be effectively applied in practice. Priority research areas will be adapted after consultation and student participation in international competitions and active participation in international conferences are foreseen.
Key words: deep learning, convolution neural networks, neural networks architecture, detection, instance segmentation, image processing, computer vision, neural network inference
more info
year 2022/2023