RxGS-pretrained

Pretrained checkpoints for the NeurIPS 2026 paper "RxGS: Receiver-Generalizable 3D Gaussian Splatting for Radio-Frequency Data Synthesis" by Kang Yang and Mani Srivastava.

📄 Paper: arXiv:2605.24290  ·  💻 Code: github.com/nesl/RxGS  ·  🗂️ Data: kyang73/RxGS-data

One RxGS model per dataset, each serving all receivers in its scene. Each checkpoint holds the Gaussian parameters and the receiver-conditioning network; optimizer state is not included.

File	Dataset	Receivers	Iterations (Stage I + II)
ble_rssi.pth	BLE RSSI	21	30k + 100k
spectrum_multirx.pth	Spatial spectrum	21	30k + 60k
csi.pth	WiFi CSI	8	30k + 100k

Usage

From the root of the code repository, with the datasets in data/:

hf download kyang73/RxGS-pretrained --include "*.pth" --local-dir pretrained
python -m scripts.inference_ble_multirx      --config arguments/configs/exp_ble_multirx_main.yaml      --checkpoint pretrained/ble_rssi.pth
python -m scripts.inference_spectrum_multirx --config arguments/configs/exp_spectrum_multirx_main.yaml --checkpoint pretrained/spectrum_multirx.pth
python -m scripts.inference_csi_multirx      --config arguments/configs/exp_csi_multirx_main.yaml      --checkpoint pretrained/csi.pth

Outputs are written to pretrained/<name>_inference/.

Citation

@inproceedings{yang2026rxgs,
  title     = {RxGS: Receiver-Generalizable 3D Gaussian Splatting for Radio-Frequency Data Synthesis},
  author    = {Yang, Kang and Srivastava, Mani},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year      = {2026}
}

License

BSD 3-Clause License.