GOOSE 3D Semantic Segmentation Challenge Label Data

This dataset consists of semantically segmented LiDAR point clouds of the GOOSE and GOOSE-Ex dataset.

The original point clouds annotations segmented all points into 64 semantic classes, but for the GOOSE 3D Semantic Segmentation Challenge on CodaBench we consolidated the point cloud data into 8 superclasses (+ other class):

category_name,label_key,hex
other,0,#A9A9A9
artificial_structures,1,#DE88DE
artificial_ground,2,#EBFF3B
natural_ground,3,#A1887F
obstacle,4,#FFC107
vehicle,5,#F44336
vegetation,6,#4CAF50
human,7,#8FB0FF
sky,8,#2196F3

The dataset contains 13006 annotated point clouds in total, stored in the .label format as is done in the SemanticKITTI dataset.

import numpy as np

# reading a .label file 
label = np.fromfile(filename, dtype=np.uint32) 
label = label.reshape((-1))

# extract the semantic and instance label IDs 
sem_label = label & 0xFFFF # semantic label in lower half 
inst_label = label >> 16 # instance id in upper half

This dataset only contains the annotations, to download the LiDAR point cloud data, please visit the download dataset page in the GOOSE dataset documentation.