Automatic detection of screen collapse by ensemble learning from images on Web screens.

Abstract

Automated Web application testing generates screenshots as test evidences. In this paper proposed, automated detection of Web screens collapse, really occurred, that difficult to find out manually from huge number of screenshots. Convolutional Neural Network (CNN) can detect them in small areas and similarity verdict by comparing two images with Random Forest (RF) classification can detect them in large areas in the images. The result verifies ensemble learning of CNN and RF effectively detecting the collapse which enables reducing work efforts of software testing.