sample_nodules() works strange

[theVmagnificient]

from radio.preprocessing.utils import get_nodules_pixel_coords, show_slices
get_nodules_pixel_coords(batch)

returns dataframe with 8 nodules

from radio.preprocessing.utils import num_of_cancerous_pixels
num_of_cancerous_pixels(batch)

returns 1 scan with 2k nodules pixels

However when I'm trying to attachsample_nodules(batch_size=None, nodule_size=(16, 32, 32), share=1.0) to the current pipeline and calling batch_crops = (dicom_dataset >> crops_sampling_pipeline).next_batch(20)

I'm getting nothing here despite I used share = 1.0 get_nodules_pixel_coords(batch_crops).head(1)

[akoryagin]

Hello, @theVmagnificient!

Thanks for the issue! It does look strange. It seems to me, though, that there is some incidental shuffle=True that changes the order of iteration through the dataset. Another possibility is that batch is not the first batch, generated by corresponding pipeline. In any case, it would very helpful if you could provide us with the full gist, that leads to this result.

Best, Alex