Croatian Academy of Engineering (HATZ)
2018
A large part of present technological achievements results from research and continuing advances in the field of artificial intelligence (AI). AI is a part of computer science that aims to create intelligent machines, capable of thinking, acting and learning like humans. It is an interdisciplinary field spanning a variety of subfields, among which machine learning (ML) and computer vision (CV) are generally regarded as core parts of AI. CV is a field that aims to give the computer visual understanding of the world from images. ML is a field of study that gives computers the ability to learn how to solve a certain task. It is particularly suited for problems that may seem relatively simple for humans, but are rather difficult to solve by using classical image processing approaches. CV and ML fields have a significant overlap where many CV problems can be solved using ML techniques. Several papers listed below present a part of CV and related ML research conducted by experts from two laboratories of the Center of Excellence for Computer Vision (CRV) at the University of Zagreb Faculty of Electrical Engineering and Computing and also by CRV collaboration researchers. The first laboratory involved is Human-oriented Technologies Laboratory (HOTLab) led by Prof. Igor S. Pandžić, Ph.D., while the second laboratory involved is Advanced Shape Reconstruction and Registration Laboratory (SHARK Lab) led by Tomislav Pribanić, Ph.D., Associate Prof. Nowadays one heavily studied ML application is certainly face analysis (FA) presented in one of the papers below. Applications of FA technologies range from marketing and entertainment to automotive industry in which, for instance, the goal is fatigue detection for vehicle driver. Another paper presented discusses two thoroughly researched CV tasks: object localization and semantic segmentation. The former attempts to find objects in the input image, where minimum bounding rectangle of the object and the associated object class are the ideal output. The latter is somewhat more detailed where each image pixel is assigned to the corresponding class label. Interesting applications can be found in traffic control systems and medical imaging. The next paper presents ML in the context of image categorization and image similarity whereby a commercial service was developed, enabling buyers of certain products to find visually similar objects of interest. The camera is the essential tool used in CV. For numerous geometry related tasks the camera requires calibration which affects many applications such as geocoding, as explained in another paper. A geometrically calibrated camera is a basis for the 3D passive and 3D active reconstruction system too. 3D scanning systems are extensively used in fashion design and development and medical applications such as human back surface analyses. The last two papers put emphasis on those two applications.