Daniel Helm
Supervisor(s): Martin Kampel
TU Wien
Abstract: The first impression of a person can decide about a positive or negative outcome in different situations. The human brain is able to get a picture of the counterpart's personality at short notice. The main aim of this paper is to comprehend how an automated system can be built to predict the Big-Five personality traits of a person. Therefore, a Convolutional Neural Network is trained on visual data features extracted from short video sequences. This paper investigates how various pre-processing methods, like face-extraction and data-augmentation, influence the predicted personality confidences. Furthermore, it explores different deep learning techniques e.g. regularization in order to improve the video-based predictions of the Big-Five personality traits. Moreover, this paper points out the complexity of developing and training a Convolutional Neural Network based system to solve a regression task. Finally, the results derived in this paper are compared to those reported by the winning teams of the ”First Impressions Challenge 2016” organized by ChaLearn Looking At People, published at the European Conference of Computer Vision 2016.
Keywords: Computer Vision, Image Processing, Multimedia
Full text: Year: 2019
