Professor Krzysztof Krawiec: Many AI applications that seemed hypothetical just a few years ago are now within our reach
Nowadays image recognition (IR) is an extremely rapidly developing area of artificial intelligence offering a wide variety of applications. It is also an area in which, apart from natural language processing, the most progress has been made in improving systems efficiency.
As a result, many AI applications that only several years ago seemed hypothetical are now within our reach. Artificial intelligence may be used in medicine, autonomous vehicles, monitoring, bioidentification and the like.
Based on literature, Internet sources (Google Scholar and Scopus databases) and author’s own experience, the author has made a list of Polish research centers and groups for which image recognition (understood as part of AI, i.e. extensive use of machine learning in practice) is an important field for basic and/or applied research. In his analysis of the state of research, the author focused mainly on centers and researchers conducting basic research, as centers using IR in various practical contexts (e.g. medicine, biology, pharmacology, quality control, monitoring, environmental protection, astronomy etc.) are too numerous to be presented in a concise form.
The following review covers mainly those people and centers dealing with IR that, to the best of author’s knowledge and according to the search performed, have adopted approaches typical for AI and machine learning. The list is in alphabetical order, sorted by names of leaders of research teams. The description of each team includes information about main research areas, selected project names and collaborators.
In view of the foregoing, the list should not be treated as an exhaustive or complete image of artificial intelligence in the field of image recognition in Poland. It is possible that some Polish IR researchers may have not been covered by the search (in particular, researchers who do not directly use AI methods and scientists who have started their career only recently).
Warsaw University of Life Sciences – SGGW, Department of Informatics, Division of Image Processing and Object Recognition
• Conventional techniques of image analysis, in particular the Hough transform.
• Applications in image detection, 3D image analysis (LIDAR images).
AGH University of Science and Technology in Cracow, Faculty of Computer Science, Electronics and Telecommunications, Department of Electronics
• Vehicle driver fatigue detection with the use of ANN.
• Road sign recognition with the use of machine learning, including ANN.
• Adaptive color image segmentation with the use of machine learning.
• Teaching and using multiple classifiers in image recognition.
Examples of research projects: Object recognition in low-quality images combining pattern classification methods and algorithms for visual information reconstruction in real time mode; Research on combining tensor methods and soft computing for pattern recognition.
Lodz University of Technology; Faculty of Electrical, Electronic, Computer and Control Engineering; Institute of Applied Computer Science
• Tissue segmentation in microscopic images.
• Blood vessel segmentation in computed tomography (CT) scanning.
• Use of deep ANN to estimate the age of trees on the basis of growth ring analysis.
• Analysis of magnetic resonance imaging (MRI) in diagnosing cancer and other diseases.
Poznan University of Technology; Faculty of Electrical Engineering; Institute of Control, Robotic and Information Engineering
• Detectors and descriptors in robotics, in particular in navigation.
• Face and eye detection with machine learning methods (cascade classifiers).
• Fast hardware implementations.
• New ANN architectures (e.g. spiking neural networks).
Collaborators: Piotr Skrzypczyński, Dominik Belter.
Example of a research project: Precise self-localization of a walking robot on rough terrain using parallel tracking and mapping.
Silesian University of Technology, Faculty of Automatic Control, Electronics and Computer Science
• Hybridization of machine learning methods (in particular support vector machines) and evolutionary algorithms.
• Applications in skin detection and segmentation algorithms, facial recognition, hand recognition, gesture recognition and sign language gesture recognition.
• Use of swarm algorithms to tune deep ANN hyperparameters.
Collaborators: Jakub Nalepa.
Example of a research project: Deep support vector machine architectures developed with the use of evolutionary algorithms.
Poznan University of Technology, Faculty of Computing, Laboratory of Intelligent Decision Support Systems
• Evolutionary algorithms used in learning image analysis programs, applied in medical diagnostics, license plate detection, generic object recognition, and recognition of objects with the use of synthetic aperture radar, SAR.
• ANN in pathomorphologic diagnoses (central nervous system tumors).
• Deep ANN for segmentation of ophthalmic imaging (fundus imaging and optical coherence tomography) and detection of anomalies in computed tomography lung screening.
Collaborators: Bartosz Wieloch, Paweł Liskowski.
Example of a research project: Interferometric Imaging Methods for Investigation of Dynamics of Biological Systems.
Wrocław University of Science and Technology, Faculty of Electronics, Department of Systems and Computer Networks
• Analysis of textures with machine learning techniques, applied in medical diagnostics (osteoporosis diagnostics).
• Basic research in the field of machine learning; main focus: multiple classifiers.
University of Lodz, Faculty of Mathematics and Computer Science; Institute of Fundamental Technological Research of the Polish Academy of Sciences
• Analysis of medical imaging (e.g. mammograms, ultrasonography), i.a. with the use of evolutionary algorithms.
University of Zielona Góra, Institute of Control and Computation Engineering
• Segmentation of pathomorphologic images in breast cancer diagnostics and application of machine learning methods in that area.
• Analysis of cytological images (pathomorphology).
AGH University of Science and Technology in Cracow, Faculty of Management, Department of Applied Computer Science
• Biometric systems.
• Medical imaging understanding.
• Cognitive computational intelligence in medical pattern semantic understanding.
Warsaw University of Technology, Faculty of Electrical Engineering, Institute of Theory of Electrical Engineering, Measurement and Information Systems, Division of Theory of Electrical Engineering and Applied Computer Science
• Shape analysis.
• Fourier and wavelet descriptors, ANN.
Systems Research Institute of the Polish Academy of Sciences (and University of Alberta, Canada)
• Application of fuzzy systems and granular computing in the image analysis and interpretation.
• New ANN architectures (granular neural networks, self-organizing polynomial neural networks, polynomial-based radial basis function neural networks).
• Application of alternative (if compared to gradient methods) learning algorithms, e.g. particle swarm optimization algorithms.
• Applications: mainly in facial recognition.
Example of a research project: Rejection in pattern recognition problem: ideas, methods, analyses.
Czestochowa University of Technology, Institute of Computational Intelligence
• New ANN architectures, including probabilistic and fuzzy networks.
• Neural and fuzzy systems.
• Multiple classifiers.
• Handwriting analysis, mainly for identification purposes.
• Image segmentation, stereo vision.
Collaborators: Rafał Scherer, Krzysztof Cpałka, Marcin Korytkowski.
Silesian University of Technology; Faculty of Automatic Control, Electronics and Computer Science; Institute of Automatic Control; Department of Exploratory Data Analysis
• Adaptive median filtering, noise reduction.
• Segmentation of images of central nervous system structures.
• Analysis of medical ultrasound images.
• Emotion recognition based on facial images.
Example of a research project: Detection and recognition of non-verbal deception markers.
University of Information Technology and Management in Rzeszow, Faculty of Applied Information Technology (also: Ohio University, Athens, USA),
• New neural networks architectures, with particular regard given to memory (e.g. neural episodic memory).
• Agent systems.
• Applications in navigation of robots.
Example of a research project: Development of robot perception mechanisms using motivated learning and self-organizing associative memory.
Lodz University of Technology, Medical Electronics Division
• Human–computer interfaces used in applications for the disabled.
• Application in quality control and in processing of biomedical signals.
AGH University of Science and Technology in Cracow, Department of Automatic Control and Biomedical Engineering
• Neural networks in IR applications.
• Knowledge discovery in IR, especially in medical applications (SPECT imaging, X-ray imaging).
Collaborators: Ewa Dudek-Dyduch, Lidia Ogiela, Marek Ogiela.
University of Social Sciences, Warsaw Campus, Faculty of Humanities (and: University of Louisville, Kentucky, USA)
• Alternative ANN architectures, models and learning algorithms.
• Extraction of features, especially unsupervised extraction of features.
• ANN unsupervised learning, especially autoencoder architectures.
Industrial and implementation research in business
In the light of progress in basic science and technological development, the share of commercial entities operating in the IR R&D area is getting increasingly bigger. However, the scope of that tendency is hard to evaluate due to limited access to information (which is kept strictly confidential by many companies for obvious reasons).
Nevertheless, according to public information, advanced research in the field of image analysis and processing is conducted by the following entities (in alphabetical order):
• Antmicro, Poznań/Wrocław, https://antmicro.com/
• Deepsense.ai, Warsaw, https://deepsense.ai/
• Future Processing, Gliwice, https://www.future-processing.com/
• Optopol Technology, Zawiercie, http://www.optopol.com.pl/
• Roche Polska, Poznań, https://www.roche.pl/
As in the case of other parts of this study, the above list is not exhaustive.
This paper is an adaptation of the study by professor Krzysztof Krawiec entitled “Status and perspectives of artificial intelligence development in the field of image recognition and computer vision” (Poznań 2018), prepared for the National Information Processing Institute.