BSc thesis - day 65

All the work is done and here it is ...

[ thesis_cover.jpg]

Now all I have to do is to wait until Friday for the presentation and then I am free! :)

The English version of the abstract goes like this:

Object categorization using Support Vector Machine with hierarchical local features

In this BSc thesis we present a computer object categorization model that combines statistically learned hierarchical local features and a linear Support Vector Machine. We describe a method for learning a database of features on a set of natural images and show how the constructed features can robustly be detected in images. A method that combines the learned hierarchy of features and the linear SVM is presented. We performed the experiments to measure the categorization success of the model based on four parameters - the level of features hierarchy, the size of features database, the size of learning set for the Support Vector Machine and the way the input data for the Support Vector Machine is constructed. The model was tested on the Caltech101 image database. In the conclusion some ideas for further development and research are given.