Seta: ambiente de realidade aumentada para o ensino de algoritmos de aprendizagem de máquina

Abstract

Augmented reality (AR) is the overlay of virtual images into images of a real environment, in real time. AR environments can be used as supplementary teaching resources for different content . This paper presents an augmented reality environment called SETA, for teaching machine learning algorithms, specifically different versions of k -means clustering algorithm and algorithms of linear and polynomial regression. Given a set of points on a whiteboard , the teacher can select which algorithms to apply and then display the results on the real image in real time, containing the plotted points . The image is captured using a webcam connected to a computer, which is the system platform. The image processing algorithm which performs segmentation and interpretation of the input images uses a global thresholding technique , mathematics morphology, labelling and Hough transform. The software was developed in Java and it was used image processing libraries (ImageJ and OpenCV) and machine learning library Weka . The tests showed that the system is able to operate satisfactorily under the lighting conditions observed in real situations , and with a refresh rate suitable for the application .