jarrelscy/radar-onnx

RADAR (ONNX, browser build)

ONNX conversion of the RADAR generalist abdominal CT model by Alibaba DAMO Academy (checkpoint_radar_pretrain.pth), split into three parts so it can run in a browser with ONNX Runtime Web. Used by radar-web (live at https://jarrelscy.github.io/radar-web/).

Research use only. Not a medical device and not for diagnosis or treatment decisions.

file	does
part_a.onnx	stem and first stage-2 conv, run on 8-slice slabs
part_b.onnx + part_b.data0/.data1	rest of the encoder, decoder to 1/4 size, token embeddings
part_c.onnx	last decoder stage, segmentation head, upsample, argmax, run on 8-slice slabs
head.bin / head.json	attention head, projections and precomputed text embeddings for 146 findings

The conversion script is tools/export.py in the radar-web repo. On the upstream demo case the outputs match the PyTorch model to within 1e-5.

Licence and attribution

Original model and weights: alibaba-damo-academy/damo-radar, CC BY-NC-SA 4.0. These converted files are released under the same licence.

@article{damo-radar-2026,
    author = {Qi Zhang and Jianpeng Zhang and Weiwei Cao and Zilin Lu and Wanxing Chang and Haonan Ding and Cao Chen and Zhi Li and Xing Xue and Sinuo Wang and Shaoteng Zhang and Yutong Xie and Yong Xia and Qi Wu and Zhongyi Shui and Xi Li and Zhilin Zheng and Yanjie Zhou and Tony C.W. Mok and Yingda Xia and Hongkan Wang and Xianghua Ye and Tao Ma and Jie Peng and Xiaoguang Wang and Jian Ding and Yuming Gao and Huazhen Ye and Yiping Liu and Dongjie Chen and Zhaomin Ni and Jianwen Ning and Wei Zhang and Jian Liu and Chaohui Yu and Shenghong Ju and Jianfeng Zhang and Wenbo Xiao and Ling Zhang and Tingbo Liang},
    title = {An expert-level generalist AI for abdominal CT diagnosis},
    journal = {Science},
    volume = {393},
    number = {6817},
    pages = {eaec6129},
    year = {2026},
    doi = {10.1126/science.aec6129}
}