Calling status on a pipeline without labels results in error

[maartenterpstra]

Hi, I try to use this library for transforming my first set of images with a known ground truth. These have no labels, I just want transformed images for use with my network. Therefore, I try to follow the DataPipeline example as follows:

>> import glob
>> import numpy
>> from PIL import Image
>> import Augmentor
>> ref_imgs = glob.glob('./reference_images/**/*.jpg')
>> imgs = glob.glob('./images/**/*.jpg')
>> collated = list(zip(ref_imgs, imgs))
>> len(collated)
130475
>> images = [[np.asarray(Image.open(y)) for y in x] for x in collated]
>> p = Augmentor.DataPipeline(images)
>> p.rotate(1, max_left_rotation=5, max_right_rotation=5)
>> p.flip_top_bottom(0.5)
>> p.zoom_random(1, percentage_area=0.5)
>> p.status()
Operations: 3
	0: RotateRange (probability=1 max_left_rotation=-5 max_right_rotation=5 )
	1: Flip (probability=0.5 top_bottom_left_right=TOP_BOTTOM )
	2: ZoomRandom (probability=1 percentage_area=0.5 randomise=False )
Images: 130475
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/local_scratch2/mterpst5/anaconda3/lib/python3.7/site-packages/Augmentor/Pipeline.py", line 798, in status
    label_pairs = sorted(set([x.label_pair for x in self.augmentor_images]))
  File "/local_scratch2/mterpst5/anaconda3/lib/python3.7/site-packages/Augmentor/Pipeline.py", line 798, in <listcomp>
    label_pairs = sorted(set([x.label_pair for x in self.augmentor_images]))
AttributeError: 'list' object has no attribute 'label_pair'

This should not happen because my data has no labels. I suppose a check should be in place to make sure labels are present.

Thanks

[Charles-Lu]

I reproduce the same bug. For now you can comment out p.status() to avoid the error. The remaining of the pipeline works perfectly.
P.S. also remember to use

augmented_images = p.sample(10)

rather than

augmented_images, labels = p.sample(10)

in following steps.

[mdbloice]

Hi @mlterpstra-umc, @Charles-Lu this is definitely a bug with the status() function, thanks to both of you for rooting it out, I will fix ASAP.

[mdbloice]

Ok so I have made a quick bug fix, this fix won't appear in the pip installed version of Augmentor until I make a version update which will happen a little later.