Generalization between different views in object recognition

with Zili Liu (UCLA)

How is recognition performance affected by different viewpoints and object categories?

Goal

Despite the large literature on the influence of viewpoint and object categories on object recognition (Tarr & Bulthoff, 1995; Bulthoff & Edelman, 1992; Biederman & Gerhardstein, 1993,1995), the extent to which object recognition can be generalized from canonical to non-canonical view and vice-versa has not been investgated enough.In this project we are investigating the degree of viewpoint generalization for different viewpoints for different levels of object categories in image-based and object-based tasks.