Classification of basic circuit components by image processing methods

Electrical and Electronics Engineers can easily recognize and distinguish circuit components when they see circuits drawn by hand on paper. However, it is difficult for computers and machines to classify hand drawn circuit components and to detect circuit components on hand drawn circuits. For this purpose, there are some studies in the literature on the classification and recognition of hand drawn circuit components using different methods. In this thesis, two different experimental studies have been conducted to classify hand drawn circuit components and to detect them on the circuit. In the first experimental study, four different Convolutional Neural Networks (CNN) models were created and hand drawn circuit components were classified. And the performances of four different CNN models have been compared. As a result of the experimental study, the method used has reached a high success rate in the classification of hand drawn circuit components. In the second experimental study, by using the CNN-based Faster Region Based Convolutional Neural Networks (Faster R-CNN) method, hand drawn circuit components were detected on many hand drawn circuits. The Faster R-CNN method, on the other hand, performed the detection of circuit components in hand drawn circuits drawn in different styles, with low loss and fast performance.