Online Image Analysis to Identify Politicians in Twitter Images (Bachelor Thesis, Finished)

Author

Florian Mohler

Description

Social media such as Twitter and Facebook are nowadays an essential part of our lives and they are often used for recreational purposes, but also increasingly more in the fields of business, science, as well as politics. The immense amount of data generated by the high number of users offers great opportunities to explore and answer questions in a scientific or political context through deeper analysis.

In this Bachelor’s thesis, we develop a system that allows matching faces, that are found in Twitter images, with a collection of reference images. For this purpose, we extract feature vectors from each of the faces, which are then stored in a database and used to compute possible matches using k-Nearest-Neighbor search. In a simple user interface, the images, matches, and reference images can be viewed. The results of the evaluation, which tested the accuracy of the face recognition methods used and their performance, suggest that the implemented methods are indeed viable for possible real-world applications. The individual components of the project are modular and allow the functionality to be extended with additional features.

Start / End Dates

2020/10/26 - 2021/01/25

Supervisors

Research Topics