GD-MAE: Generative Decoder for MAE Pre-training on LiDAR Point Clouds

TL;DR

GD-MAE introduces a simple generative decoder for 3D point cloud pre-training that effectively restores masked geometric information, achieving state-of-the-art results with less latency and high robustness across large-scale benchmarks.

Contribution

The paper proposes a novel generative decoder paradigm for 3D MAE pre-training that simplifies the architecture and enhances performance and flexibility over previous complex methods.

Findings

Outperforms existing methods on Waymo, KITTI, and ONCE benchmarks.

Achieves comparable accuracy with only 20% labeled data on Waymo.

Reduces decoding latency by over 88% compared to traditional approaches.

Abstract

Despite the tremendous progress of Masked Autoencoders (MAE) in developing vision tasks such as image and video, exploring MAE in large-scale 3D point clouds remains challenging due to the inherent irregularity. In contrast to previous 3D MAE frameworks, which either design a complex decoder to infer masked information from maintained regions or adopt sophisticated masking strategies, we instead propose a much simpler paradigm. The core idea is to apply a \textbf{G}enerative \textbf{D}ecoder for MAE (GD-MAE) to automatically merges the surrounding context to restore the masked geometric knowledge in a hierarchical fusion manner. In doing so, our approach is free from introducing the heuristic design of decoders and enjoys the flexibility of exploring various masking strategies. The corresponding part costs less than \textbf{12\%} latency compared with conventional methods, while achieving better performance. We demonstrate the efficacy of the proposed method on several large-scale benchmarks: Waymo, KITTI, and ONCE. Consistent improvement on downstream detection tasks illustrates strong robustness and generalization capability. Not only our method reveals state-of-the-art results, but remarkably, we achieve comparable accuracy even with \textbf{20\%} of the labeled data on the Waymo dataset. Code will be released at https://github.com/Nightmare-n/GD-MAE.