chrischoy/FCGF
A 3D fully-convolutional neural network that extracts compact geometric features from point clouds for registration and correspondence tasks.
Not currently ranked — collecting fresh signals.
star history
FCGF extracts geometric features from 3D scans or point clouds using a single-pass fully-convolutional network, achieving state-of-the-art accuracy without preprocessing. The model produces compact 32-dimensional features and is optimized for speed with MinkowskiEngine sparse tensor operations on GPU. It is designed for applications like 3D registration, reconstruction, and tracking in both indoor and outdoor scenes.
Frequently asked
- What is chrischoy/FCGF?
- A 3D fully-convolutional neural network that extracts compact geometric features from point clouds for registration and correspondence tasks.
- Is FCGF open source?
- Yes — chrischoy/FCGF is open source, released under the MIT license.
- What language is FCGF written in?
- chrischoy/FCGF is primarily written in Python.
- How popular is FCGF?
- chrischoy/FCGF has 733 stars on GitHub.
- Where can I find FCGF?
- chrischoy/FCGF is on GitHub at https://github.com/chrischoy/FCGF.