Running multi-input model with Python client

Healthcare Clara Train: AI-Assisted Annotation SDK
#1

Hi, how can I use the “client.segmentation()” function (from client_api.py) for nvidia models with multiple inputs such as the “segmentation_mri_brain_tumors_br16_full” model?

I unsuccessfully tried something like:

mask = client.segmentation(model = 'segmentation_mri_brain_tumors_br16_full',
image_in = ['./Brats18/Brats18_TCIA02_322_1_t1ce.nii.gz',
'./Brats18/Brats18_TCIA02_322_1_t1.nii.gz',
'./Brats18/Brats18_TCIA02_322_1_t2.nii.gz',
'./Brats18/Brats18_TCIA02_322_1_flair.nii.gz'],
image_out = './Brats18/Brats18_TCIA02_322_1_seg.nii.gz')

My idea was if to follow the same order as found in “seg_brats18_datalist_0.json” under the config folder of the ngc model.

#2

Hi,

That is a good try.
However for AIAA now it does not take multiple inputs.

To run the model,
You can try either

  1. merge these volumes to a single 4D volume and put it in image_in

  2. download 4D brain volumes from medical decathon: http://medicaldecathlon.com/#tasks
    then put it there.

thanks