Emergent patterns are patterns that are common across many images in an image data set. Infants are apparently able to discover such emergent patterns and eventually form object categories, largely unsupervised. It is interesting to see if unsupervised computational approaches can lead to similar outcomes on large image sets. In this talk I will discuss two algorithms used for obtaining these emergent patterns, based on data mining and graph based approaches.
The results show that interesting patterns emerge, including, but not limited to, object categories.
Last modified: Tuesday, 23-Sep-2014 14:23:03 NZST
This page is maintained by the seminar list administrator.