SC²S Colloquium - October 26, 2017

From Sccswiki
Jump to navigation Jump to search
Date: October 26, 2017
Room: 00.08.053
Time: 3:00 pm, s.t.

Tim Waegemans: Image Classification with Geometrically Aware Sparse Grids

In this bachelor thesis a reduction of sparse grids is presented, to deal with the large number of dimensions of images. The reduction is based on the geometric structure of image data. The information of neighboring pixels is weighted more than pixels at different sides of the image. Additionally the use of Sparse Grid Density Estimation and Modified Linear Basis functions helps reaching the high dimensionality. An efficient evaluation of the modified grid is presented. The method is applied to different image datasets, showing promising results.