Computer detection of circuit elements in electrical diagrams

U ovom radu opisan je postupak klasifikacije slika elemenata električne sheme. Na početku je napravljen kratak uvod u strojno učenje i računalni vid. Nakon toga, opisan je teorijski dio dvaju najčešće korištenih algoritama za klasifikaciju slika, a to su stroj potpornih vektora i konvolucijske neuronske mreže. Oba algoritma su korištena za izgradnju modela te su izgrađeni modeli međusobno uspoređeni. Također, detaljno je opisan proces prikupljanja slika te njihova obrada. Programsko rješenje je napisano u programskom jeziku Python uz korištenje biblioteka OpenCV i TensorFlow. This thesis describes the procedure of image classification of electrical circuit elements. A Brief introduction of machine learning and computer vision was made at the beginning. After that, the theoretical part of the two most commonly used algorithms for image classification is described, and they are called the support vector machine and the convolutional neural network. Both algorithms were used to build models which were compared to each other. As well, the process of collecting images and their processing is described in detail. The programming solution was written in Python programming language using the OpenCV and TensorFlow libraries.