zju3dv/LoFTR
LoFTR uses a transformer architecture to match local features between image pairs for pose estimation and 3D vision tasks.
Not currently ranked — collecting fresh signals.
star history
LoFTR implements a detector-free approach to local feature matching using transformer networks. Rather than relying on traditional feature detectors, it processes images jointly through a transformer to establish correspondences between image pairs. The method demonstrates strong performance on pose estimation benchmarks and has been integrated with Huggingface Spaces for accessibility.
Frequently asked
- What is zju3dv/LoFTR?
- LoFTR uses a transformer architecture to match local features between image pairs for pose estimation and 3D vision tasks.
- Is LoFTR open source?
- Yes — zju3dv/LoFTR is open source, released under the Apache-2.0 license.
- What language is LoFTR written in?
- zju3dv/LoFTR is primarily written in Jupyter Notebook.
- How popular is LoFTR?
- zju3dv/LoFTR has 2.9k stars on GitHub.
- Where can I find LoFTR?
- zju3dv/LoFTR is on GitHub at https://github.com/zju3dv/LoFTR.