II INTERNATIONAL SCIENTIFIC CONFERENCE OF YOUNG RESEARCHERS
590
Qafqaz University
18-19 April 2014, Baku, Azerbaijan
information with object interaction like grasping, shaking, pushing , pressing and etc. by means of that feature vector
formalized and object with similar feature vector combined under same class. The same approach also can be observed in
“Learning affordances for categorizing objects and their properties” [6] article that interaction of robot and object can lead
to manifestation of object perceptual properties, after that robot will be able to predict effects of actions on object from
which perceptual features extracted. The authors of “Using object affordances to improve object recognition” [7] article
additionally propose to use motor information of object rather that only using visual information. In this experiment for each
training object its associated grasp created (it’s the visual mapping of object according to its affordance) in other words
visuomotor map formed. The article show that, the presence of motor information increased effect of recognition, even if
this information in test data do not include actual motor information, it latter can be inferred from visuomotor map.
In my opinion the best approach is “Multimodal object categorization by a robot” [4] because for beginning our aim is
predict the class of object rather than object itself, so if the class of object is recognized then robot will be able to behave
with objects taking in account the affordance of class, which in my opinion will give correct output.
Refferences
1) (The Ecological Approach to Visual Perception, Houghton Mifflin, Boston, 1979.)
2) “Distinctive image features from scale-invariant key points” (D.G. Lowe International Journal of Computer Vision
60 (2) (2004) 91–110.)
3) “Efficient object recognition capabilities in online robots: from a statistical to a neural-network classifier” by P.
Sanz, R. Marin, J. Sanchez, (IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews 35
(1) (2005) 87–96.)
4) “Multimodal object categorization by a robot” (T. Nakamura, T. Nagai, N. Iwahashi in: IEEE/RSJ International
Conference on Intelligent Robots and Systems, IROS 2007, 2007, pp. 2415–2420.)
5) “Grounding semantic categories in behavioral interactions: experiments with 100 objects” (J. Sinapov, C. Schenck,
K. Staley, V. Sukhoy, A. Stoytchev,Robotics and Autonomous Systems (0) (2012).)
6) “Learning affordances for categorizing objects and their properties” (N. Dag, I. Atil, S. Kalkan, E. Sahin in: 20th
International Conference on Pattern Recognition, ICPR 2010, 2010, pp. 3089–3092.)
7) “Using object affordances to improve object recognition” (C. Castellini, T. Tommasi, N. Noceti, F. Odone, B.
Caputo, IEEE Transactions on Autonomous Mental Development 3 (3) (2011) 207–215.)
Dostları ilə paylaş: |