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

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.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
3D Point Cloud Classification ModelNet40 PointNeXt Overall Accuracy 94.0 # 21
Mean Accuracy 91.1 # 18
Number of params 4.5M # 96
FLOPs 6.5G # 1
Semantic Segmentation S3DIS PointNeXt-XL Mean IoU 74.9 # 11
mAcc 83.0 # 10
oAcc 90.3 # 10
FLOPs 84.8G # 1
Number of params 41.6M # 52
Params (M) 41.6 # 1
3D Semantic Segmentation S3DIS PointNext mIoU (6-Fold) 74.9 # 3
mIoU (Area-5) 70.5 # 3
Semantic Segmentation S3DIS PointNeXt-L Mean IoU 73.9 # 15
mAcc 82.2 # 15
oAcc 89.9 # 14
FLOPs 15.2G # 1
Number of params 7.1M # 42
Params (M) 7.1 # 11
Semantic Segmentation S3DIS Area5 PointNeXt mIoU 71.1 # 22
oAcc 91.0 # 15
mAcc 77.2 # 19
Number of params 41.6M # 53
3D Point Cloud Classification ScanObjectNN PointNeXt Overall Accuracy 88.2 # 26
Mean Accuracy 86.8 # 10
Number of params 1.4M # 50
FLOPs 1.64G # 1
3D Part Segmentation ShapeNet-Part PointNeXt Class Average IoU 85.2 # 4
Instance Average IoU 87.1 # 3

Methods