Abstract
There are some classification methods that generate nuances in the final accuracy caused by objects positioning, framing and damage. These occurrences may result in a drop of accuracy in computer vision systems that were trained with structured static datasets and are intended to be used in day-to-day applications in which the images are not always as organized as the trained dataset, like some biometric classification systems such as iris and fingerprint. In this regard, this paper presents six image datasets processed with different methods to help researchers analyze the impact of object positioning, framing and damage in their taxonomies.