In the dataset description, it says that "The training set annotations are provided through either two ways: human-labeling and camera-radar fusion (CRF) algorithm". Can we have access to which of these annotations are labeled by humans? I think this may help analyze a new annotation method, and maintain the human-labeled annotations while changing the CRF-labeled ones.
Posted by: zangwei @ Jan. 31, 2021, 4:52 p.m.This is not supposed to be provided to participants. Well, you can design an algorithm to distinguish which part is manually annotated if you need to.
Posted by: ywang26 @ Feb. 1, 2021, 5:53 a.m.