P3P: Pseudo-3D Pre-training for Scaling 3D Voxel-based Masked Autoencoders

TL;DR

This paper introduces P3P, a self-supervised pre-training framework that efficiently incorporates large-scale image data into 3D voxel-based models, improving 3D perception tasks like classification and segmentation.

Contribution

The paper proposes a novel linear-time tokenizer and a new 3D reconstruction target to enhance 3D pre-training with diverse and large-scale data.

Findings

Achieves state-of-the-art results in 3D classification

Improves performance in few-shot learning

Enhances 3D segmentation accuracy

Abstract

3D pre-training is crucial to 3D perception tasks. Nevertheless, limited by the difficulties in collecting clean and complete 3D data, 3D pre-training has persistently faced data scaling challenges. In this work, we introduce a novel self-supervised pre-training framework that incorporates millions of images into 3D pre-training corpora by leveraging a large depth estimation model. New pre-training corpora encounter new challenges in representation ability and embedding efficiency of models. Previous pre-training methods rely on farthest point sampling and k-nearest neighbors to embed a fixed number of 3D tokens. However, these approaches prove inadequate when it comes to embedding millions of samples that feature a diverse range of point numbers, spanning from 1,000 to 100,000. In contrast, we propose a tokenizer with linear-time complexity, which enables the efficient embedding of a flexible number of tokens. Accordingly, a new 3D reconstruction target is proposed to cooperate with our 3D tokenizer. Our method achieves state-of-the-art performance in 3D classification, few-shot learning, and 3D segmentation. Code is available at https://github.com/XuechaoChen/P3P-MAE.