Hi-3DOVS & Hi-LERF

We construct two 3D hierarchical semantics datasets, named Hi-LERF and Hi-3DOVS.We annotate 40 images from 4 scenes based on the LERF and 60 images from 10 scenes based on the 3D-OVS. Each scene contains 3 hierarchical levels of mask sets, with each set containing approximately 10 corresponding annotations.