Rust inference binary for the brainssl segmentation and connectome models (ONNX Runtime).
261
A small Rust binary that runs both brainssl models without Python, on ONNX Runtime: tumour segmentation from four-contrast MRI, and autism classification from ROI time series.
Source and documentation: https://github.com/mahdisabetkish/brainssl (folder rust/).
Models: https://huggingface.co/mahdisabetkish
docker pull mahdisabetkish/brainssl-infer
docker run --rm -v "$PWD:/data" mahdisabetkish/brainssl-infer segment --model /models/segmentation --input /data/scan.nii.gz --output /data/scan_seg.nii.gz
The image is about 40 MB. It does the preprocessing around each network (midline search, symmetric crop, sliding windows, mirror averaging for segmentation; Ledoit-Wolf, matrix logarithm and window averaging for connectomes) and was checked against the Python pipeline. Research use only, not a medical device.
Content type
Image
Digest
sha256:083050858…
Size
78.1 MB
Last updated
10 days ago
docker pull mahdisabetkish/brainssl-infer