Write subvolumes into .h5 format

Course Q&A AI For Medicine AI for Medical Diagnosis
1

Hi

How exactly do you write the extracted subvolumes into .h5 format?
I have been trying on this but seems I’m not doing it the right way.
And also is there any visualising tool to check the .h5 file is correct or not?

2

Hi @Shay,

A few years ago I wrote code to make a .h5 file for some images, hope this helps. Also, I used jupyter notebook to visualise the files.

import numpy as np
import h5py
import glob

from matplotlib.image import imread

image = {}
I=0
img_mask = '/Users/mubsi/Desktop/Total2/*.jpg'
img_names = glob.glob(img_mask)
for fn in img_names:
    print(i, fn)
    image[i] = imread(fn)
    i = i + 1

with h5py.File('/Users/mubsi/Desktop/Test.h5', 'w' ) as hdf:
    hdf.create_dataset('trainSet_x', shape=(47,64,64))
    dset = hdf['trainSet_x']
    for x in range(0, 47):
        dset[x] = image[x]

Cheers,
Mubsi

1 Like
3

Hi @Mubsi,
Thanks for answering my question.
I thought the write-in method I used was wrong. But I recently found that when I generate the training volume on google colab, the data in the .h5 file became null. It’s pretty weird… the file size was 11.5MB and after training it became 800KB