xingyul/flownet3d
FlowNet3D is a deep neural network that estimates 3D scene flow (point motion) from point cloud data in an end-to-end manner.
Not currently ranked — collecting fresh signals.
star history
This repository implements FlowNet3D, a deep learning model published at CVPR 2019, designed to estimate scene flow directly from 3D point clouds. The network learns hierarchical features of point clouds and flow embeddings representing point motions through novel learning layers for point sets. It is trained on synthetic FlyingThings3D data and evaluated on KITTI Lidar scans, demonstrating generalization to real-world data.
Frequently asked
- What is xingyul/flownet3d?
- FlowNet3D is a deep neural network that estimates 3D scene flow (point motion) from point cloud data in an end-to-end manner.
- Is flownet3d open source?
- Yes — xingyul/flownet3d is open source, released under the MIT license.
- What language is flownet3d written in?
- xingyul/flownet3d is primarily written in Python.
- How popular is flownet3d?
- xingyul/flownet3d has 403 stars on GitHub.
- Where can I find flownet3d?
- xingyul/flownet3d is on GitHub at https://github.com/xingyul/flownet3d.