Human-in-the-Loop Optimization for Product Design with Gaze Data

The purpose of this work is to search for a possible link between some intrinsic features of a three-dimensional model and the emotions it experiences in people who observe it. Specifically, we are interested in looking for what these features are and how it affects aesthetic perception with the goal of using these inputs to improve the final design of a product. In this thesis we will start describing the techniques and methods used to conduct the experiments and to perform the analysis, moving then to a brief review of related research in the field. The core part of the thesis will explain how we manipulated gaze data through surrogate-based optimization and statistical analysis in order to enhance the aesthetic perception of a 3D model. We then analyze the performance of our model on an already existing dataset with a case study approach. Finally, we will conclude the document by analyzing the outcomes and indicating possible applications and future improvements of the work. The results obtained as a result of the application of the above techniques are that there is no statistically significant relationship between the intrinsic characteristics of the model and the interest aroused in the subjects subjected to the experiment.