PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies

By Guocheng Qian, Houwen Peng, Jinjie Mai, Hasan Hammoud, Mohamed Elhoseiny, Bernard Ghanem. Neural Information Processing Systems, 2022

1,153 citations recorded by the source corpus.

PointNet++ is one of the most influential neural architectures for point cloud understanding. Although the accuracy of PointNet++ has been largely surpassed by recent networks such as PointMLP and Point Transformer, we find that a large portion of the performance gain is due to improved training strategies, i.e. data augmentation and optimization techniques, and increased model sizes rather than architectural innovations. Thus, the full potential of PointNet++ has yet to be explored. In this work, we revisit the classical PointNet++ through a systematic study of model training and scaling strategies, and offer two major contributions. First, we propose a set of improved training strategies that significantly improve PointNet++ performance. For example, we show that, without any change in architecture, the overall accuracy (OA) of PointNet++ on ScanObjectNN object classification can be raised from 77.9% to 86.1%, even outperforming state-of-the-art PointMLP. Second, we introduce an inverted residual bottleneck design and separable MLPs into PointNet++ to enable efficient and effective model scaling and propose PointNeXt, the next version of PointNets. PointNeXt can be flexibly scaled up and outperforms state-of-the-art methods on both 3D classification and segmentation tasks. For classification, PointNeXt reaches an overall accuracy of 87.7 on ScanObjectNN, surpassing PointMLP by 2.3%, while being 10x faster in inference. For semantic segmentation, PointNeXt establishes a new state-of-the-art performance with 74.9% mean IoU on S3DIS (6-fold cross-validation), being superior to the recent Point Transformer. The code and models are available at https://github.com/guochengqian/pointnext.

Subfields: Computational Mechanics, Computer Vision and Pattern Recognition, Geology

View Paper: https://www.semanticscholar.org/paper/4ea12bc8dab90e3b7b01b19148ff37ad325f54cd

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

1. Zhikang Zou
2. Peng Gao
3. P. Heng
4. Renrui Zhang
5. Xiang Bai
6. Xin Zhou
7. Xiao Tan
8. Xingkui Zhu

In Computer Vision and Pattern Recognition

1. Peng Gao
2. Renrui Zhang
3. Hongsheng Li
4. Jiaming Han
5. Ziyu Guo
6. P. Heng
7. Ke Chen
8. Zhikang Zou

In Geology

1. Peng Gao
2. P. Heng
3. Renrui Zhang
4. Hongsheng Li
5. Ziyu Guo

Citations in the corpus

• Point-Bind & Point-LLM: Aligning Point Cloud with Multi-modality for 3D Understanding, Generation, and Instruction Following (2023, methodology): “In the last two columns, we utilize different 3D encoders in Point-Bind, i.e., Point-BERT [86], PointNeXt [57], and I2P-MAE [92].”
• PointMamba: A Simple State Space Model for Point Cloud Analysis (2024, background): “To address the issues, there has been rapid progress in deep learning-based methods [38, 33, 44, 67], consistently pushing the performance to the new record.”
• 3D Diffuser Actor: Policy Diffusion with 3D Scene Representations (2024, neutral): “4) PolarNet [8], a 3D policy that computes dense point representations for the robot workspace using a PointNext backbone [57].”