为点云自监督学习赋予掩码自编码器显式对比特性
Bringing Masked Autoencoders Explicit Contrastive Properties for Point Cloud Self-Supervised Learning
AI总结：
本文提出 Point-CMAE，通过双重随机掩码、共享编码器与双解码器重建，并约束共同掩码 token 的特征一致性，在 MAE 点云预训练中引入显式对比学习，提升下游性能。
摘要：
图像领域中用于 Vision Transformers（ViTs）的对比学习（CL）已经取得了可媲美传统卷积骨干网络 CL 的性能。然而，在使用 ViTs 进行三维点云预训练时，掩码自编码器（MAE）建模仍占主导地位。这就引出了一个问题：我们能否结合两者的优点？为回答该问题，我们首先通过实验验证，将基于 MAE 的点云预训练与标准对比学习范式相结合，即使经过精心设计，也可能导致性能下降。为克服这一局限，我们利用 MAE 固有的对比特性，将 CL 重新引入基于 MAE 的点云预训练范式。具体而言，我们不依赖图像领域常用的大量数据增强，而是对输入 token 进行两次随机掩码，以生成对比输入对。随后，采用权重共享编码器和两个结构相同的解码器执行掩码 token 重建。此外，我们提出，对于同时被两种掩码遮住的输入 token，其重建特征应尽可能相似。这自然地在基于 MAE 的生成式预训练范式中建立了显式对比约束，由此形成我们提出的方法 Point-CMAE。因此，与 MAE 基线相比，Point-CMAE 有效提升了表征质量和迁移性能。在分类、部件分割和少样本学习等多种下游应用上的实验评估表明，在标准 ViTs 和单模态设置下，我们的框架优于最先进技术。源代码和训练模型可在 https://github.com/Amazingren/Point-CMAE 获取。
英文摘要：
Contrastive learning (CL) for Vision Transformers (ViTs) in image domains has achieved performance comparable to CL for traditional convolutional backbones. However, in 3D point cloud pretraining with ViTs, masked autoencoder (MAE) modeling remains dominant. This raises the question: Can we take the best of both worlds? To answer this question, we first empirically validate that integrating MAE-based point cloud pre-training with the standard contrastive learning paradigm, even with meticulous design, can lead to a decrease in performance. To address this limitation, we reintroduce CL into the MAE-based point cloud pre-training paradigm by leveraging the inherent contrastive properties of MAE. Specifically, rather than relying on extensive data augmentation as commonly used in the image domain, we randomly mask the input tokens twice to generate contrastive input pairs. Subsequently, a weight-sharing encoder and two identically structured decoders are utilized to perform masked token reconstruction. Additionally, we propose that for an input token masked by both masks simultaneously, the reconstructed features should be as similar as possible. This naturally establishes an explicit contrastive constraint within the generative MAE-based pre-training paradigm, resulting in our proposed method, Point-CMAE. Consequently, Point-CMAE effectively enhances the representation quality and transfer performance compared to its MAE counterpart. Experimental evaluations across various downstream applications, including classification, part segmentation, and few-shot learning, demonstrate the efficacy of our framework in surpassing state-of-the-art techniques under standard ViTs and single-modal settings. The source code and trained models are available at: https://github.com/Amazingren/Point-CMAE.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
