How is the version 2.0 dataset's ontology different from version 1.0?
linzhiqiu opened this issue · comments
In particular, do we have a detailed analysis of how the ontology evolved from version 1.0 to 2.0? i.e. what is the relationship between classes in version 2.0 (v2.0) and classes in version 1.0 (v1.0)?
For an example, these are the subclasses under the superclass "traffic-light" in version 2.0:
Do these fine-grained classes belong to the standalone "traffic-light" class in version 1.0? In other words -- the original "traffic-light" class (of v1.0) breaks down into these subclasses in v2.0?
For another example, these are classes under superclass "traffic-sign" in v2.0:
The original v1.0 only has "front" and "back" under the superclass "traffic-sign". Therefore, are the rest of these classes in v2.0 part of the original "front" and "back" class in v1.0?