Icon Semantic Classification

In this thesis, you primarily focus on creating a dataset and developing a machine-learning algorithm to classify icons found in various user interfaces.

Completed Bachelor Thesis

The aim of this thesis is to determine how the semantic classification of icons can be implemented in an algorithm. The algorithm receives a section of a user interface (UI) as input, which potentially contains an icon, and is intended to output the semantic class of the icon. For this purpose, a literature search in the field of image recognition, classification, reverse engineering of UIs and traffic sign recognition is carried out followed by implementation of a solution using the convolutional neural network (CNN). In order to be able to train and evaluate the CNN, building a dataset is necessary in which icons are assigned to their semantic classes. To achieve this, a dataset from a related work [1] is expanded using web icons from various available sources.

Applicants should have proficient knowledge in programming and basic knowledge in machine learning or deep learning approaches. If you are interested in this topic, please contact Ramin Hedeshy or Raphael Menges.

[1]. Thomas F. Liu, Mark Craft, Jason Situ, Ersin Yumer, Radomir Mech, and Ranjitha
Kumar. Learning design semantics for mobile apps. In Proceedings of the 31st
Annual ACM Symposium on User Interface Software and Technology, UIST ’18, pages
569–579, New York, NY, USA, 2018. ACM.

Supervisors

To the top of the page