FRTSearch: Fast Radio Transient Search

FRTSearch is an end-to-end deep learning framework for detecting and characterizing Fast Radio Transients (FRTs), including: Pulsars, Rotating Radio Transients (RRATs) and Fast Radio Bursts (FRBs).

Model Info

Item	Value
Backbone	HRNet-W32
Input	256 × 8192 (freq × time)
Size	400 MB
Formats	.fits (PSRFITS), .fil (Filterbank)
Bit Depth	1/2/4/8/32-bit

QUICK START

from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="waterfall109/FRTSearch",
    filename="models/hrnet_epoch_36.pth"
)

Or download directly from Files and versions.

TEST SAMPLES

This repository includes 5 test samples from 2 different telescopes to demonstrate cross-facility performance:

Telescope	FRB	DM (pc cm⁻³)
FAST	20121102, 20180301, 20201124	565, 420, 525
ASKAP	20180119, 20180212	400, 168

CITATION

@article{Zhang_2026,
  author  = {Zhang, Bin and Wang, Yabiao and Xie, Xiaoyao and You, Shanping and Yu, Xuhong and Li, Qiuhua and Li, Hongwei and Du, Shaowen and Miao, Chenchen and Zhou, Dengke and Fang, Jianhua and Wu, Jiafu and Wang, Pei and Li, Di},
  title   = {FRTSearch: Unified Detection and Parameter Inference of Fast Radio Transients Using Instance Segmentation},
  journal = {The Astrophysical Journal Supplement Series},
  volume  = {284},
  number  = {2},
  pages   = {58},
  year    = {2026},
  doi     = {10.3847/1538-4365/ae5e76},
}

Test Sample References

When using the test samples, please also cite the original observations:

• FAST samples: Guo et al. (2025)
• SKA samples: Shannon et al. (2018)

License & Acknowledgments

GPL-2.0 | Based on MMDetection & PRESTO

_{Exploring the dynamic universe with AI 🌌📡 | Issues}