yuxumin/PoinTr
Transformer-based deep learning model for completing incomplete 3D point clouds from partial observations.
Not currently ranked — collecting fresh signals.
star history
PoinTr converts point clouds into sequences of point proxies with positional embeddings and uses a transformer encoder-decoder architecture to generate complete 3D shapes. The work also introduces more challenging benchmarks (ShapeNet-55/34) for point cloud completion research. The repository contains PyTorch implementations of PoinTr and its follow-up variant AdaPoinTr.
Frequently asked
- What is yuxumin/PoinTr?
- Transformer-based deep learning model for completing incomplete 3D point clouds from partial observations.
- Is PoinTr open source?
- Yes — yuxumin/PoinTr is open source, released under the MIT license.
- What language is PoinTr written in?
- yuxumin/PoinTr is primarily written in Python.
- How popular is PoinTr?
- yuxumin/PoinTr has 854 stars on GitHub.
- Where can I find PoinTr?
- yuxumin/PoinTr is on GitHub at https://github.com/yuxumin/PoinTr.