3D point cloud analysis has drawn a lot of research
attention due to its wide applications. However, collecting
massive labelled 3D point cloud data is both timeconsuming
and labor-intensive. This calls for data-efficient
learning methods. In this work we propose PointDisc,
a point discriminative learning method to leverage selfsupervisions
for data-efficient 3D point cloud classification
and segmentation. PointDisc imposes a novel point discrimination
loss on the middle and global level features
produced by the backbone network. This point discrimination
loss enforces learned features to be consistent with
points belonging to the corresponding local shape region
and inconsistent with randomly sampled noisy points. We
conduct extensive experiments on 3D object classification,
3D semantic and part segmentation, showing the benefits
of PointDisc for data-efficient learning. Detailed analysis
demonstrate that PointDisc learns unsupervised features
that well capture local and global geometry.
License type:
Publisher Copyright
Funding Info:
This research / project is supported by the A*STAR - MTC Young Individual Research Grant
Grant Reference no. : M21K3c0130
This research / project is supported by the A*STAR - Career Development Award
Grant Reference no. : 202D8243