Unsupervised Point Cloud Pre-training via Occlusion Completion

TLDR

The paper proposes an unsupervised pre‑training method for point clouds that reconstructs occluded points. The method masks occluded points in a camera view, trains an encoder‑decoder to reconstruct them, and then uses the encoder weights to initialize downstream point‑cloud models. Pre‑training on a single dataset improves accuracy on diverse downstream tasks, surpassing prior methods, and yields features with wide minima, high transformation invariance, and activations strongly correlated with part labels. Code and data are available at https://github.com/hansen7/OcCo.

Abstract

We describe a simple pre-training approach for point clouds. It works in three steps: 1. Mask all points occluded in a camera view; 2. Learn an encoder-decoder model to reconstruct the occluded points; 3. Use the encoder weights as initialisation for downstream point cloud tasks. We find that even when we pre-train on a single dataset (ModelNet40), this method improves accuracy across different datasets and encoders, on a wide range of downstream tasks. Specifically, we show that our method outperforms previous pre-training methods in object classification, and both part-based and semantic segmentation tasks. We study the pre-trained features and find that they lead to wide downstream minima, have high transformation invariance, and have activations that are highly correlated with part labels. Code and data are available at: https://github.com/hansen7/OcCo