zahrag / bioscan-1m Goto Github PK

Hello,

Thanks a lot for making the dataset available. I'm really looking forward to trying to work with it.

I have just started looking at the data and am having trouble reading the images from cropped_256.hdf5. Perhaps this is just due to my own unfamiliarity with hdf5, but it seems like the images are 1 dimensional and that dimension does not seem necessarily divisible by 256.

I am reading the images in as follows:

import numpy as np

with h5py.File('cropped_256.hdf5','r+') as hdf5:
    img = np.array(hdf5['bioscan_dataset']['3652597.jpg'])
    print(img.shape)
    print(img.shape / 256)

and I get the following output:

(7633,)
29.81640625

I have tried with a few different images and I get similar results. For now I'll use the zip files so this probably isn't a barrier for many, but if I am doing something wrong in the reading of the hdf5 data, maybe a how-to in the readme would be a good idea.

Thanks again for the dataset!