PointCNN
A deep learning framework for feature learning directly from point clouds using X-Conv operations, achieving state-of-the-art results in classification and segmentation.
What is PointCNN?
PointCNN is a deep learning framework for processing 3D point cloud data, introducing the X-Conv operator to perform convolution-like operations directly on irregular point sets. It solves the problem of effective feature extraction from unordered point clouds, enabling tasks like 3D object classification, part segmentation, and scene understanding without requiring voxelization or other intermediate representations.
Researchers and developers working in 3D computer vision, geometric deep learning, and point cloud processing, particularly those needing state-of-the-art performance on benchmarks like ModelNet40, ScanNet, and S3DIS.
PointCNN offers a general and simple framework that achieves record-breaking accuracy on multiple point cloud benchmarks by learning spatial transformations that enable effective convolution on irregular points, avoiding the information loss and computational cost of voxel-based methods.
Overview
PointCNN: Convolution On X-Transformed Points (NeurIPS 2018)
Use Cases
Best For
- 3D object classification from point cloud data
Related Projects
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
