05: Supervised Segmentation Pipeline issue

[thomasnilsson]

The following piece of code

from pipe_utils import flatten_step
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import RobustScaler
digit_pipe = Pipeline([('Flatten', flatten_step),
                       ('Normalize', RobustScaler())])
digit_pipe.fit(img_data)

show_pipe(digit_pipe, img_data)
show_pipe(digit_pipe, img_data, show_hist=True)

Produces the following error

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-39-fb8d4dbf3178> in <module>()
      6 digit_pipe.fit(img_data)
      7 
----> 8 show_pipe(digit_pipe, img_data)
      9 show_pipe(digit_pipe, img_data, show_hist=True)

~/GitHub/QBI-ethz/Quantitative-Big-Imaging-2019-master/Lectures/pipe_utils.py in show_pipe(pipe, in_data, show_hist)
     52                     last_data = step_op.predict(last_data)
     53 
---> 54         display_data(c_ax, last_data, show_hist)
     55         c_ax.set_title('Step {} {}\n{}'.format(i, last_data.shape, step_name))
     56         c_ax.axis('on')

~/GitHub/QBI-ethz/Quantitative-Big-Imaging-2019-master/Lectures/pipe_utils.py in display_data(in_ax, raw_data, show_hist)
     33         else:
     34             n_stack = np.stack([(x-x.mean())/x.std() for x in in_data], 0)
---> 35             in_ax.imshow(montage2d(n_stack))
     36 
     37 

TypeError: 'module' object is not callable

[kmader]

@thomasnilsson You have a different version of skimage (which renamed montage2d to montage).

A temporary fix is to change the second line of pipe_utils.py from from skimage.util import montage as montage2d to from skimage.util.montage import montage2d and it should work.

If you are comfortable using pip and upgrading packages then you can also run pip install scikit-image==0.14.2 and everything will work without making any changes

[kmader]

Did that work for you or is there still an issue?

[thomasnilsson]

pip install scikit-image==0.14.2 worked wonders, thank you for the help!