onemore1/onmi-fmri repository overview

⁠Omni-fMRI: A Universal Atlas-Free fMRI Foundation Model

This repository contains the official implementation of Omni-fMRI: A Universal Atlas-Free fMRI Foundation Model, which introduces a dynamic patching mechanism that significantly reduces computational costs while preserving informative spatial structures.

More information please follows https://github.com/OneMore1/Omni-fMRI⁠

⁠How to use Docker

⁠Option A: Build Locally

Run from the Omni-fMRI repository root:

docker build -t omnifmri:local .

Start the container:

docker run --gpus all --rm -it \
  --ipc=host \
  -v "$PWD":/workspace \
  -v /path/to/data_root:/data \
  -v /path/to/outputs:/outputs \
  omnifmri:local \
  bash

Inside the container, edit config paths to container paths, for example:

data:
  data_root: /data

Then run:

sh scripts/pretrain.sh
# or
sh scripts/finetune.sh

⁠Option B: Use The Published Image

docker pull onemore1/onmi-fmri:latest

Run from the Omni-fMRI repository root:

docker run --gpus all --rm -it \
  --ipc=host \
  -v "$PWD":/workspace \
  -v /path/to/data_root:/data \
  -v /path/to/outputs:/outputs \
  onemore1/onmi-fmri:latest \
  bash

⁠Model Checkpoints

Our pre-trained model weights can be found in Huggingface. https://huggingface.co/OneMore1/Omni-fMRI⁠

⁠Model Docker

Docker can be found in Dockerhub. https://hub.docker.com/r/onemore1/onmi-fmri⁠

⁠Citation

Citations and discussions are welcome.

@article{wang2026omni, title={Omni-fMRI: A Universal Atlas-Free fMRI Foundation Model}, author={Wang, Mo and Ye, Wenhao and Xia, Junfeng and Zhang, Junxiang and Pan, Xuanye and Xu, Minghao and Deng, Haotian and Wen, Hongkai and Liu, Quanying}, journal={arXiv preprint arXiv:2601.23090}, year={2026} }