Palestrante
Descrição
Visual attention is a complex cognitive process where individuals select the most relevant and informative stimuli to understand the environment, thus reducing information processing time and improving interaction skills. Computational modeling approaches to visual attention propose that saliency maps are built from bottom-up visual features in the scene context, highly determining the attentional focus trajectory. Furthermore, these maps may be complemented by top-down cognitive aspects of information processing such as pattern recognition and the visual task nature. The majority of saliency maps are obtained through computer vision feature extraction and some of these techniques utilises complex networks characterization of images for saliency map extraction, but we have few studies with a complex networks approach into built saliency maps. (1) In our study we modeled the attentional focus trajectory as a complex network derived from saliency maps of atificial scenes. Salient regions in the scene corresponded to nodes whose connections corresponded to attentional focus displacement probabilities. Connections were weighted both by bottom-up visual feature interactions and top-down pattern recognition and task demands. We studied how topological and dynamic measurement (2) of these complex networks can lead to prediction of attentional focus displacement and compared our approach to traditional saliency models and visual attention literature.
Referências
1 BORJI, A.; ITTI, L. State-of-the-art in visual attention modeling. IEEE Transactions on Pattern Analysis and Machine Intelligence, v. 35, n. 1, p. 185-207, Jan. 2013.
2 COSTA, L. F. et al. Characterization of complex networks: a survey of measurements. Advances in Physics, v. 56, n. 1, p. 167-242, Feb. 2007.
| Subárea | Física Computacional e Simulações Numéricas |
|---|---|
| Apresentação do trabalho acadêmico para o público geral | Não |
