This repository hosts code and related resources for research on Learning-based Geometry Point Cloud Transmission.

This repository includes two major components:

The first folder contains the full codebase of SEPT, a conference paper which focuses on the transmission of small point cloud datasets generated by downsampling the point clouds in ShapeNet.

Please refer to the SEPT repository for full details.

The second folder contains the implementation of the OTA-MetaNeRF framework of the paper Over-the-Air Learning-based Geometry Point Cloud Transmission, this scheme considers transmitting point clouds using neural network weights, and adopts Meta learning to accelerate encoding process.

This work adopts the principle of Coinpp and apply it for efficient (real-time) and effective 3d point cloud transmission. Unlike SEPT for small-scale point clouds, this scheme is generalizable to all point cloud inputs.

Please refer to the readme.md file in the OTA-MetaNeRF folder.

If you find this work useful in your research, please cite:

@article{ota-pct,
  title={Over-the-Air Learning-based Geometry Point Cloud Transmission},
  author={Bian, Chenghong and Shao, Yulin and Gunduz, Deniz},
  journal={IEEE Journal on Selected Areas in Communications},
  year={2025},
}