Machine vision group, the research team from the Institute of Automation and Informatics, has long been involved in the development and implementation of robust computer vision systems. They are also engaged in the development of methods and instrumentation for sustainable machine vision. Both research directions are thus focused on digitalization and Industry 4.0.
Deep learning has significantly advanced the field of computer vision, attracting researchers and developers from various sectors, including healthcare, the food industry, and industrial manufacturing. The proliferation of tools has enabled the creation of hundreds of deep neural network applications daily. These applications are typically trained using publicly available datasets and often demonstrate high accuracy in controlled environments. However, they frequently falter in real-world applications, highlighting a lack of robustness in computer vision systems.
To address this issue, the machine vision group has been dedicated to exploring this phenomenon. Their primary focus includes developing methods for handling imbalanced datasets, enhancing data augmentation techniques, and designing effective data collection methodologies. With the support of the machine vision lab, these proposed techniques are rigorously tested on real-world industrial manufacturing problems and beyond.
Machine vision is a significant driving force behind Industry 4.0. With the advent of affordable and powerful hardware, automating visual inspection tasks on an unprecedented scale has become possible. However, the proliferation of these applications has led to a growing energy burden in industrial production. Addressing this issue requires the development of data processing methods that maintain high accuracy and processing speed while being energy-efficient. Additionally, reducing energy consumption through improved instrumentation is also crucial. The machine vision group is dedicated to both of these approaches.