Automatic testing of graphical user interfaces

Automatic testing is a part of the development process when creating GUIs.Traditionally, this is a time consuming process that is performed both by the supplier, and the company ordering the GUI.By using image processing, this process can be performed dynamically without having to specify rules manually.In this thesis, two methods for identifying interactive areas have been studied.These methods are identification using edge detection as well as convolutional neural networks.The evaluation was done by comparing how accurate these methods were when faced with GUIs of varying complexity.The neural network approach was found to perform better in all aspects, which led to a testing framework that can navigate through a novel GUI automatically.The result of searching through a GUI is presented in the form of a graph which can be used to study GUI complexity.