Nachricht

Am Mittwoch, den 15. Juni 2016 hält Herr David Roling um 16:30 Uhr im Raum P 1.4.17 einen Vortrag über seine Masterarbeit mit dem Titel:

Text Mining Methods for Semantic Annotation of View-Based Object Representations

Abstract:

This work focuses on the development and evaluation of a system for automatic semantic annotation of view-based object representations. In a first step, given an object representation by one ore more views of an object, an image search engine is used to find web pages containing visually similar images. The texts from those web pages typically contain information about the visually similar images and can therefore be used to extract semantic labels describing the views of an object. For this purpose, a central objective of this work is the selection, implementation and adaptation of a text mining method able to extract such semantic labels. While representative works from the literature most often use automatic image annotation algorithms on large scale and manually created databases with images together with corresponding semantic labels, the system developed in this work does not need such a prior database. The system is evaluated with a range of images from objects of different difficulties and compared to existing approaches.