This Bachelor’s thesis delves into the recognition and classification of objects in graphical user interfaces (GUIs), a vital area for automation, testing, and migration tasks. Utilizing modern techniques like Computer Vision and Deep Neural Networks, the study achieves a significant improvement in prediction quality, boasting an F1 score of 0.524 at an IoU threshold of >0.9. The research addresses three key challenges: independent verification of existing results, applicability to Web User Interfaces, and the need for higher absolute prediction quality. The work aims to resolve these issues through a replication study, exploring state-of-the-art techniques, reproducing published results, and improving prediction quality, all while evaluating the findings using appropriate datasets and metrics.