Alert button

Self-Supervised Point Cloud Representation Learning with Occlusion Auto-Encoder

Add code
Alert button
Mar 26, 2022
Figure 1 for Self-Supervised Point Cloud Representation Learning with Occlusion Auto-Encoder
Figure 2 for Self-Supervised Point Cloud Representation Learning with Occlusion Auto-Encoder
Figure 3 for Self-Supervised Point Cloud Representation Learning with Occlusion Auto-Encoder
Figure 4 for Self-Supervised Point Cloud Representation Learning with Occlusion Auto-Encoder

Share this with someone who'll enjoy it:

Learning representations for point clouds is an important task in 3D computer vision, especially without manually annotated supervision. Previous methods usually take the common aid from auto-encoders to establish the self-supervision by reconstructing the input itself. However, the existing self-reconstruction based auto-encoders merely focus on the global shapes, and ignore the hierarchical context between the local and global geometries, which is a crucial supervision for 3D representation learning. To resolve this issue, we present a novel self-supervised point cloud representation learning framework, named 3D Occlusion Auto-Encoder (3D-OAE). Our key idea is to randomly occlude some local patches of the input point cloud and establish the supervision via recovering the occluded patches using the remaining visible ones. Specifically, we design an encoder for learning the features of visible local patches, and a decoder for leveraging these features to predict the occluded patches. In contrast with previous methods, our 3D-OAE can remove a large proportion of patches and predict them only with a small number of visible patches, which enable us to significantly accelerate training and yield a nontrivial self-supervisory performance. The trained encoder can be further transferred to various downstream tasks. We demonstrate our superior performances over the state-of-the-art methods in different discriminant and generative applications under widely used benchmarks.

Share this with someone who'll enjoy it: