Recognition of Transitive Actions with Hierarchical Neural Network Learning

Proceedings of the 25th International Conference on Artificial Neural Networks, pages 472--479, doi:10.1007/978-3-319-44781-0_56 - Sep 2016.
Associated documents : Mici_ICANN_2016.pdf [867Ko]   http://dx.doi.org/10.1007/978-3-319-44781-0_56
The recognition of actions that involve the use of objects has remained a challenging task. In this paper, we present a hierarchical self-organizing neural architecture for learning to recognize transitive actions from RGB-D videos. We process separately body poses extracted from depth map sequences and object features from RGB images. These cues are subsequently integrated to learn action-object mappings in a self-organized manner in order to overcome the visual ambiguities introduced by the processing of body postures alone. Experimental results on a dataset of daily actions show that the integration of action-object pairs significantly increases classification performance.

 

@InProceedings\{MPW16,
  author       = "Mici, Luiza and Parisi, German I. and Wermter, Stefan",
  title        = "Recognition of Transitive Actions with Hierarchical Neural Network Learning",
  booktitle    = "Proceedings of the 25th International Conference on Artificial Neural Networks",
  pages        = "472--479",
  month        = "Sep",
  year         = "2016",
  address      = "Barcelona, ES",
  doi          = "10.1007/978-3-319-44781-0_56",
  url          = "https://www2.informatik.uni-hamburg.de/wtm/publications/2016/MPW16/Mici_ICANN_2016.pdf"
}

» Luiza Mici
» German I. Parisi
» Stefan Wermter