Medforj Brain T1w 3d

--- language: - en license: cc-by-nc-4.0 tags: - image-generation - brain - 3d pipeline_tag: unconditional-image-generation --- # medforj-brain-t1w-3d `medforj-brain-t1w-3d` has a set of 3D DDPMs with various prediction types for 3D T1w MRI brain generation, as described in the paper [Diffusion-Driven Generation of Minimally Preprocessed Brain MRI](https://huggingface.co/papers/2510.26834). For more information, please see [our GitHub page.](https://github.com/piksl-research/medforj/) # Usage We provide a reference implementation of `medforj-brain-t1w-3d`, as well as sampling code, in a dedicated [github repository](https://github.com/piksl-research/medforj/). Developers and creatives looking to build on top of `medforj-brain-t1w-3d` are encouraged to use this as a starting point. Here is a quick example from the GitHub repository to sample an image from the pre-trained weights: ```bash # First, clone the repository and install dependencies cd $HOME && git clone https://github.com/piksl-research/medforj.git cd $HOME/medforj conda create -n medforj python==3.11 conda activate medforj pip install . # Then, sample an image from the pre-trained weights python demo.py \ --out-fpath {MY_IMG.nii.gz} \ --weight-fpath {WEIGHTS.pt} \ --prediction-type {sample/velocity/flow} \ --gpu-id 0 ``` Choose the `prediction_type` according to the pre-trained weights used. The EMA weights are available on HuggingFace at this repository. # License This model falls under the CC 4.0 BY-NC license. # Citation If you find our code or models useful in your work, please consider citing them as: ```bibtex @misc{medforj-t1, title={Diffusion-Driven Generation of Minimally Preprocessed Brain MRI}, author={Samuel W. Remedios and Aaron Carass and Jerry L. Prince and Blake E. Dewey and others}, year={2025}, eprint={2510.26834}, url={https://huggingface.co/papers/2510.26834}, } ```

Source context: 0 downloads · 0 likes · Pipeline unconditional-image-generation · Repo piksl-research/medforj-brain-t1w-3d