tflearn has dependency with PIL about tflearn HOT 9 CLOSED

Hey, can you please send me the error message?
PIL is used for some pre-processing and utility functions in TFLearn.

from tflearn.

i test it on jupyter notebook.

import tflearn
import numpy as np
from PIL import Image

img_path = "...img_path..."
Image.open(img_path)

Error log:

---------------------------------------------------------------------------
IOError                                   Traceback (most recent call last)
/usr/local/lib/python2.7/dist-packages/IPython/core/formatters.pyc in __call__(self, obj)
    341             method = _safe_get_formatter_method(obj, self.print_method)
    342             if method is not None:
--> 343                 return method()
    344             return None
    345         else:

/usr/local/lib/python2.7/dist-packages/PIL/Image.pyc in _repr_png_(self)
    617         from io import BytesIO
    618         b = BytesIO()
--> 619         self.save(b, 'PNG')
    620         return b.getvalue()
    621 

/usr/local/lib/python2.7/dist-packages/PIL/Image.pyc in save(self, fp, format, **params)
   1656 
   1657         # may mutate self!
-> 1658         self.load()
   1659 
   1660         save_all = False

/usr/local/lib/python2.7/dist-packages/PIL/ImageFile.pyc in load(self)
    233         if not self.map and not LOAD_TRUNCATED_IMAGES and e < 0:
    234             # still raised if decoder fails to return anything
--> 235             raise_ioerror(e)
    236 
    237         # post processing

/usr/local/lib/python2.7/dist-packages/PIL/ImageFile.pyc in raise_ioerror(error)
     57     if not message:
     58         message = "decoder error %d" % error
---> 59     raise IOError(message + " when reading image file")
     60 
     61 

IOError: broken data stream when reading image file

Out[1] : <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=84x256 at 0x7FDF4DA8EC10>

import numpy as np
from PIL import Image

img_path = "...img_path..."
Image.open(img_path)

Out[1] : display image

from tflearn.

I think this is an issue with PIL, maybe try to update pillow.

from tflearn.

pillow version is 3.2.0.(latest)
so.. i reinstall pillow but, result is same.

from tflearn.

You can try to

sudo apt-get install libjpeg8-dev

And reinstall pillow

from tflearn.

libjpeg8-dev is already the newest version.

I also reinstall libjpeg8-dev and Pillow.
result is same.

My desktop's os is Ubuntu 14.04 LTS.
I will debug this problem.
When i figure it out, i'll tell you.

from tflearn.

Ok thanks!

from tflearn.

This problem is tensorflow's Issue.

TensorFlow_0.8.0 somehow breaks PIL ? #2101
scikit-image/scikit-image#2000

Simple solution is

from PIL import Image
import tensorflow as tf

or
Use TensorFlow 0.7.1 version. (It is the 0.8.0 version's problem)

from tflearn.

I can confirm this error when running alexnet.py with TensorFlow 0.8:
Starting to parse images... Traceback (most recent call last): File "alexnet.py", line 27, in <module> X, Y = oxflower17.load_data(one_hot=True) File "/usr/local/lib/python2.7/dist-packages/tflearn/datasets/oxflower17.py", line 34, in load_data categorical_Y=one_hot) File "/usr/local/lib/python2.7/dist-packages/tflearn/data_utils.py", line 227, in build_image_dataset_from_dir X, Y = image_dirs_to_samples(directory, resize, convert_gray, filetypes) File "/usr/local/lib/python2.7/dist-packages/tflearn/data_utils.py", line 210, in image_dirs_to_samples samples[i] = resize_image(samples[i], resize[0], resize[1]) File "/usr/local/lib/python2.7/dist-packages/tflearn/data_utils.py", line 184, in resize_image img = in_image.resize((new_width, new_height), resize_mode) File "/usr/local/lib/python2.7/dist-packages/PIL/Image.py", line 1538, in resize self.load() File "/usr/local/lib/python2.7/dist-packages/PIL/ImageFile.py", line 235, in load raise_ioerror(e) File "/usr/local/lib/python2.7/dist-packages/PIL/ImageFile.py", line 59, in raise_ioerror raise IOError(message + " when reading image file") IOError: broken data stream when reading image file
And import PIL before TensorFlow fixed this error.

from tflearn.