10.22. Clara FastIO Variable Passthrough - Nifti to FastIO Variable Conveter Operator
This application is NOT for medical use.
This application reads a Nifti file in the expected payload input folder (/input
), and assigns the values in the Nifti file to a
FastIO variable named segmentation
, while also storing the shape of the array in the Nifti file in the FastIO variable segmentation_shape
.
The operator checks the dynamically changeable dimensions of segmentation
as specified in the pipeline (e.g. if the shape
in the pipeline takes the form [1, -1, -1, -1]
it means that the last 3 dimensions in the shape are changeable at operator runtime however the array must have one “channel”). The shape (and therefore memory allocation) of segmentation
is therefore updatable using the FastIO Variable API (e.g. using segmentation_out.update_shape
which will update the type metadata and automatically allocate if all shape
entries are not -1
).
The variable segmentation_shape
is, on the other hand, fixed in size and must contain 4 elements.
The latter variable is not necessary to store shape information, since once segmentation
is fully specified shape information will directly be available from it. Here, however, we use segmentation_shape
as an exercise to use a fixed-size array.
This application, in the form of a Docker container, expects an input folder (/input
by default),
which must be mapped to a host folder. The following file is expected in the folder:
An image file in Nifti format.
FastIO variable array in shared memory.
The directories in the container are shown below.
/app_nifti_to_fastio
├── Dockerfile
├── main.py
└── requirements.txt
The license to this container is available and can be pulled as part of the procedure described above or obtained from the Clara Deploy SDK. By pulling and using the container, you accept the terms and conditions of these licenses.
Release Notes, the Getting Started Guide, and the SDK itself are available at the NVIDIA Developer forum.
For answers to any questions you may have about this release, visit the NVIDIA Devtalk forum.