Function img_to_array() gives error for JPEG on Windows

[ivan-ushakov]

Hello,

I'm getting error when I use img_to_array() function for particular JPEG file. Here is stack trace:

Traceback (most recent call last):
  File "file.py", line 41, in <module>
    tf.app.run()
  File "C:\Users\...\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\platform\app.py", line 48, in run
    _sys.exit(main(_sys.argv[:1] + flags_passthrough))
  File "file.py", line 38, in main
    print('label: ' + x.name + ', files: ' + str(load_label(x)))
  File "file.py", line 28, in load_label
    array = preprocessing.image.img_to_array(image)
  File "C:\Users\...\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\contrib\keras\python\keras\preprocessing\image.py", line 331, in img_to_array
    x = np.asarray(img, dtype=K.floatx())
  File "C:\Users\...\AppData\Local\Programs\Python\Python35\lib\site-packages\numpy\core\numeric.py", line 531, in asarray
    return array(a, dtype, copy=False, order=order)
TypeError: float() argument must be a string or a number, not 'JpegImageFile'

Image was loaded with load_img() from same package. I found that if I try to convert image directly with NumPy and without dtype parameter like np.asarray(image) then it works.

I am using tensorflow-gpu 1.1.0 on Windows 8.1, cuda8, python3.5, pillow (for PIL).

[michaelisard]

Does it work if you just do

img = pil_image.open(...)
x = np.as_array(img, dtype=float32)

i.e. just using pillow and np? If not, I guess you may have to follow up with those projects.