PAConv: Position Adaptive Convolution with Dynamic Kernel Assembling on Point Clouds

By Mutian Xu, Runyu Ding, Hengshuang Zhao, Xiaojuan Qi. Computer Vision and Pattern Recognition, 2021

557 citations recorded by the source corpus.

We introduce Position Adaptive Convolution (PAConv), a generic convolution operation for 3D point cloud processing. The key of PAConv is to construct the convolution kernel by dynamically assembling basic weight matrices stored in Weight Bank, where the coefficients of these weight matrices are self-adaptively learned from point positions through ScoreNet. In this way, the kernel is built in a data-driven manner, endowing PAConv with more flexibility than 2D convolutions to better handle the irregular and unordered point cloud data. Besides, the complexity of the learning process is reduced by combining weight matrices instead of brutally predicting kernels from point positions.Furthermore, different from the existing point convolution operators whose network architectures are often heavily engineered, we integrate our PAConv into classical MLP-based point cloud pipelines without changing network configurations. Even built on simple networks, our method still approaches or even surpasses the state-of-the-art models, and significantly improves baseline performance on both classification and segmentation tasks, yet with decent efficiency. Thorough ablation studies and visualizations are provided to understand PAConv. Code is released on https://github.com/CVMI-Lab/PAConv.

Subfields: Computational Mechanics, Computer Graphics and Computer-Aided Design, Geology

View Paper: https://www.semanticscholar.org/paper/4e6c86c08eda3937afe9e6bc298c7e03fba05352

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. Li Jiang
2. Hengshuang Zhao
3. Can Qin
4. Ranga Rodrigo
5. Xu Ma
6. Jiaya Jia
7. Isuru Dissanayake
8. Mohamed Afham

In Computer Graphics and Computer-Aided Design

1. Jiaya Jia
2. Hengshuang Zhao
3. Yun Raymond Fu
4. Xiaojuan Qi

In Geology

1. Li Jiang
2. Hengshuang Zhao
3. Can Qin
4. Xu Ma
5. Jiaya Jia
6. Xin Lai
7. Haoxi Ran
8. Jianhui Liu

Citations in the corpus

• Stratified Transformer for 3D Point Cloud Segmentation (2022, methodology): “…82.3 85.2 SubSparseCNN [15] 83.3 86.0 PointCNN [22] 84.6 86.1 PointConv [52] 82.8 85.7 Point2Sequence [23] - 85.2 PVCNN [27] - 86.2 RS-CNN [25] 84.0 86.2 KPConv [41] 85.0 86.2 InterpCNN [30] 84.0 86.3 DensePoint [24] 84.2 86.4 PAConv [54] 84 Point Transformer by a large margin in category mIoU.”
• CrossPoint: Self-Supervised Cross-Modal Contrastive Learning for 3D Point Cloud Understanding (2022, background): “Since then, numerous advancements have been made in point cloud based tasks such as 3D object classification [27, 29, 39, 49, 55, 62, 70], 3D object detection [27, 32, 37, 69] and 3D point cloud synthesis [1, 57].”
• Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Framework (2022, methodology): “Owing to the local point representation (and multi-scale information), PointNet++ exhibits promising results and has been the cornerstone of modern point cloud methods (Wang et al., 2019; Fan et al., 2021; Xu et al., 2021a).”