Key points are not available for this paper at this time.
Several recent works have explored the benefits of providing more detailed annotations for object recognition. These annotations provide information beyond object names, and allow a detector to reason and describe individual instances in plain English. However, by demanding more specific details from annotators, new difficulties arise, such as stronger language dependencies and limited annotator attention. In this work, we present the challenges of constructing such a detailed dataset, and discuss why the benefits of using this data outweigh the difficulties of collecting it.
Endres et al. (Tue,) studied this question.