Browsing by Subject "Object region"

Now showing 1 - 2 of 2

Open Access
KiMPA: A kinematics-based method for polygon approximation
(Springer, Berlin, Heidelberg, 2002) Şaykol, Ediz; Gülesir, Gürcan; Güdükbay, Uğur; Ulusoy, Özgür
In different types of information systems, such as multimedia information systems and geographic information systems, object-based information is represented via polygons corresponding to the boundaries of object regions. In many applications, the polygons have large number of vertices and edges, thus a way of representing the polygons with less number of vertices and edges is developed. This approach, called polygon approximation, or polygon simplification, is basically motivated with the difficulties faced in processing polygons with large number of vertices. Besides, large memory usage and disk requirements, and the possibility of having relatively more noise can also be considered as the reasons for polygon simplification. In this paper, a kinematics-based method for polygon approximation is proposed. The vertices of polygons are simplified according to the velocities and accelerations of the vertices with respect to the centroid of the polygon. Another property of the proposed method is that the user may set the number of vertices to be in the approximated polygon, and may hierarchically simplify the output. The approximation method is demonstrated through the experiments based on a set of polygonal objects. © Springer-Verlag Berlin Heidelberg 2002.
Open Access
On recognizing actions in still images via multiple features
(Springer, Berlin, Heidelberg, 2012) Şener, Fadime; Bas, C.; Ikizler-Cinbis, N.
We propose a multi-cue based approach for recognizing human actions in still images, where relevant object regions are discovered and utilized in a weakly supervised manner. Our approach does not require any explicitly trained object detector or part/attribute annotation. Instead, a multiple instance learning approach is used over sets of object hypotheses in order to represent objects relevant to the actions. We test our method on the extensive Stanford 40 Actions dataset [1] and achieve significant performance gain compared to the state-of-the-art. Our results show that using multiple object hypotheses within multiple instance learning is effective for human action recognition in still images and such an object representation is suitable for using in conjunction with other visual features. © 2012 Springer-Verlag.