Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Framework

Abstract

Point cloud analysis is challenging due to irregularity and unordered data structure. To capture the 3D geometries, prior works mainly rely on exploring sophisticated local geometric extractors using convolution, graph, or attention mechanisms. These methods, however, incur unfavorable latency during inference, and the performance saturates over the past few years. In this paper, we present a novel perspective on this task. We notice that detailed local geometrical information probably is not the key to point cloud analysis -- we introduce a pure residual MLP network, called PointMLP, which integrates no sophisticated local geometrical extractors but still performs very competitively. Equipped with a proposed lightweight geometric affine module, PointMLP delivers the new state-of-the-art on multiple datasets. On the real-world ScanObjectNN dataset, our method even surpasses the prior best method by 3.3% accuracy. We emphasize that PointMLP achieves this strong performance without any sophisticated operations, hence leading to a superior inference speed. Compared to most recent CurveNet, PointMLP trains 2x faster, tests 7x faster, and is more accurate on ModelNet40 benchmark. We hope our PointMLP may help the community towards a better understanding of point cloud analysis. The code is available at https://github.com/ma-xu/pointMLP-pytorch.

Accepted by ICLR 2022. Codes are made publically available at https://github.com/ma-xu/pointMLP-pytorch; updated some errors

Cited by in corpus (14)

• Radar-Camera Fusion for Object Detection and Semantic Segmentation in Autonomous Driving: A Comprehensive Review
• Deep Learning-based 3D Point Cloud Classification: A Systematic Survey and Outlook
• Point Cloud Classification Using Content-based Transformer via Clustering in Feature Space
• Advances of Machine Learning in Materials Science: Ideas and Techniques
• PointDifformer: Robust Point Cloud Registration With Neural Diffusion and Transformer
• A Closer Look at Few-Shot 3D Point Cloud Classification
• Tinto: Multisensor Benchmark for 3D Hyperspectral Point Cloud Segmentation in the Geosciences
• TTPOINT: A Tensorized Point Cloud Network for Lightweight Action Recognition with Event Cameras
• Regress Before Construct: Regress Autoencoder for Point Cloud Self-supervised Learning
• iBA: Backdoor Attack on 3D Point Cloud via Reconstructing Itself
• Beyond First Impressions: Integrating Joint Multi-modal Cues for Comprehensive 3D Representation
• PointSmile: Point Self-supervised Learning via Curriculum Mutual Information
• PosDiffNet: Positional Neural Diffusion for Point Cloud Registration in a Large Field of View with Perturbations
• Self-supervised Learning of Rotation-invariant 3D Point Set Features using Transformer and its Self-distillation