vanilla-roll is volume renderer using array-api as backend.
VTK is one of the most excellent volume renderers in the world. It has good performance and many features. But I think that the installation process is not easy. So vanilla-roll is motivated to solve this problem.
- IO
- MRA
- NIFTI
- DICOM
- NRRD
- Rendering Algorithm
- Sampling
- Shear-Warp
- Raycast
- Rendering Mode
- MIP
- MinP
- Average
- VolumeRendering
- Ambient
- Shading
- Backend
- numpy
- pytorch
- cupy
- jax
- numba
$ pip install vanilla-roll
vanilla-roll supports following extras
- torch
- dicom
- mha
- nifti
Code
import urllib.request
from pathlib import Path
from tempfile import TemporaryDirectory
import numpy as np
import skimage.io
import vanilla_roll as vr
# from A high-resolution 7-Tesla fMRI dataset from complex natural stimulation with an audio movie
# https://www.openfmri.org/dataset/ds000113/
MRA_FILE_URL = "https://s3.amazonaws.com/openneuro/ds000113/ds000113_unrevisioned/uncompressed/sub003/angio/angio001.nii.gz" # noqa: E501
def fetch_mra_volume() -> vr.volume.Volume:
with TemporaryDirectory() as tmpdir:
mra_file = Path(tmpdir) / "mra.nii.gz"
urllib.request.urlretrieve(MRA_FILE_URL, mra_file)
return vr.io.read_nifti(mra_file)
def save_result(ret: vr.rendering.types.RenderingResult, path: str):
img_array = vr.rendering.convert_image_to_array(ret.image)
skimage.io.imsave(path, np.from_dlpack(img_array)) # type: ignore
def main():
volume = fetch_mra_volume()
ret = vr.render(volume, mode=vr.rendering.mode.MIP())
save_result(ret, f"result.png")
if __name__ == "__main__":
main()
Output
If you need more exmplaes, please check the examples.
$ poetry self add 'poethepoet[poetry_plugin]'
$ poetry poe install-all-extras
$ poetry poe test
$ poetry poe check