Professorship of Applied Sciences Computer Vision & Data Science
The processing of image data (photos and videos) is one of the biggest challenges, as it is estimated that 80% of all data in the world is processed in this way. As a nationally recognised expert centre in the field of pattern recognition and image processing, the Computer Vision & Data Science professorship has been contributing to the anchoring of artificial intelligence in the Netherlands for over 25 years.
Computer Vision & Data Science is about solving automation problems in the field of visual inspections. Problems in the field of image acquisition, image processing and recognition of patterns in that image information with techniques from artificial intelligence such as deep learning and machine learning. These include quality control, automatic position and orientation determination, disease detection, defect measurement and product sorting.
The strength of the professorship is that it has both knowledge and equipment for the entire chain of illumination, cameras, optics, set-up, vision algorithms, deep learning algorithms and their implementation in existing software systems.
The four focus areas of the Computer Vision & Data Science professorship's applied research are: high-performance image acquisition, applied computer vision, state-of-the-art data science and education. The application is broad, from precision agriculture to medical. The goal and the approach are the same: to work from everyone's own strengths, in cooperation with students, staff and business experts, on state-of-the-art practice-oriented research projects in order to develop new knowledge in the field of computer vision and data science.
The Computer Vision & Data Science professorship is therefore at the cutting edge of science and practice. Our challenge is to bring (the latest) scientific knowledge in the field of image data analysis into practice. Current research themes we are working on, the translation of which to practical application still has many challenges and possibilities, are: hyperspectral imaging, anomaly detection, object tracking and explainable AI. Since the professorship works on the development of algorithms, artificial intelligence, to make the application suitable and affordable for businesses and organisations, the nature of this research is innovative and can often be done in a subsidised project.
A lot of GPU power is needed for all the experiments that the students and the team carry out during their research. Therefore, we keep updating our facilities with powerful hardware. In addition to our supercomputer called Deep Frisian, we have expanded our server network with even more computing power, bringing our capacity to 576GB of GPU memory. We combine our super computers with deep learning software algorithms that can identify visual patterns in all types of data. This is a powerful combination and the possibilities are endless.
The professorship works on automating visual inspections, such as inspecting fields with smart civilian drones and inspecting of ships on corrosion. For its research the professorship has a wide range of state-of-the-art cameras at its disposal. For example, hyper-spectral cameras to take accurate measurements in light that is not visible to humans.
Seam leakage' after an operation is a worrying problem. In the project LapVas, the Computer Vision & Data Science professorship searched for a measuring instrument that maps the quality of the intestinal wall tissue. With the aid of the measuring instrument, the micro(blood) circulation of the tissue is made visible. This enables the operating surgeon to make a better decision. The aim is that fewer intensive care admissions will be necessary, the patient will recover more quickly and fewer recovery operations will be needed. Various tests have been carried out with the measuring instrument.
Besides the MCL in Leeuwarden, the Biomedical Photonic Imaging department of the University of Twente is also actively involved in the LapVas study. The project is part of the broader research project OK of the Future by innovation agency LIMIS, which investigates innovations in healthcare.
The recycling of plastics is a well-known social challenge. The Computer Vision & Data Science professorship works together with the Circular Plastics professorship on this issue. If plastic can be sorted accurately enough, the recycling chain will work more effectively and it will no longer be necessary to burn plastic. The central question is: How can this sorting be improved using techniques from computer vision and data science?
In this project, a hyper-spectral camera is used that is sensitive to Short Wave Infrared (SWIR). In this part of the electromagnetic spectrum, it is possible to distinguish various polymers on the basis of chain length (and therefore type, PE, PP, PVC, etc.).
By using the latest techniques from artificial intelligence (e.g. deep learning), recycling systems can be trained in a data-driven way. With this approach, not only the spectral properties of polymers are included in the sorting, but also the morphological properties. In the first phase of the research, promising results were obtained that demonstrate the added value of using deep learning in the context of polymer recycling.