A question about the labels

[luna9722]

Hello,
I have a question about the boundary labels, I tried the prprocessing file and tried np.unique for one of the labels and got [0 255] A i not supposed to get 3 elements one for the background, one for the cell and one for the boundary ? to encode them later for the multi class classification for unet 3 classes ??
I would be grateful if can someone answer my question.

[jccaicedo]

You are correct, boundary labels are supposed to be 3 values. Can you visualize the resulting image and post it here? Also, can you give a reference to the specific notebook and block of code where you think the error is happening?

Thank you!

[luna9722]

It is not about an error that I found, I am just trying to understand the annotations and your code.
I took an image from the masks and tried this code from the preprocessing file in your repo.

`annot=skimage.io.imread('/content/drive/MyDrive/BBBC039/masks/IXMtest_A02_s1_w1051DAA7C-7042-435F-99F0-1E847D9B42CB.png')
print(annot.shape)
total_objects=0
if len(annot.shape) == 3:
annot = annot[:,:,0]

annot = skimage.morphology.label(annot)
total_objects += len(np.unique(annot)) - 1

annot = skimage.morphology.remove_small_objects(annot, min_size=25)

boundaries = skimage.segmentation.find_boundaries(annot)

for k in range(2, 2, 2):
boundaries = skimage.morphology.binary_dilation(boundaries)

label_binary = np.zeros((annot.shape + (3,)))

label_binary[(annot == 0) & (boundaries == 0), 0] = 1
label_binary[(annot != 0) & (boundaries == 0), 1] = 1
label_binary[boundaries == 1, 2] = 1

skimage.io.imsave('/content/imgg0.png', label_binary)

print("Total objects: ",total_objects)
imm=skimage.io.imread('/content/imgg0.png')
print(imm.shape)
print(np.unique(imm))
plt.imshow(imm)
`

ANd this is what i got

and I found it weird to get [0 255] knowing that we have 3 classes. Can you explain it to me ? Thank you in advance.