GPL3D

GPL3D DataSet from Instance-incremental Scene Graph Generation from Real-world Point Clouds via Normalizing Flows

Data process of GPL3D

Step 1. Download the Paris-Lille-3D

Step 2. Split each scene into different parts

Step 3. Split each part into different objects.

Class Labels are in Class Labels

Scene Graphs recording the object relationships are in DataSet

Class Labels

Object Class Labels:

road
sidewalk
island
Vegetal_ground
building
other_natural
tree
potted_plant
signage
bollard
trash_can
barrier
pedestrian
car
unclassified

Relationship Class Labels:

adjacent to
on
close to
supported by
leaning against
spatial proximity
behind
in front of
part of

DataSet

Lille with unclassified objects
Paris with unclassified objects
Lille without unclassified objects
Paris without unclassified objects

Environments

Ubuntu 16.04
Python 3.7
sklearn 1.0
plyfile 0.7.4