skimage >= 0.16, rescale need to set multichannel=True

[pureexe]

In skimage >= 0.16, When use skimage.transform.rescale need to set parameter multichannel=True to rescale the image
if it isn't set it will crash when input image larger than 1000 pixel in width or height because output of rescale will be 1 channel instead of 3 channels.

To reproduce bug:

1. set environment which is install skimage version 0.16 or newer.
2. input this image to demo.py with --isDlib=True
3. boom!

Tracekback from my PC.

Traceback (most recent call last):
  File "demo.py", line 168, in <module>
    main(parser.parse_args())
  File "demo.py", line 56, in main
    pos = prn.process(image) # use dlib to detect face
  File "/home/pakkapon/PRNet/api.py", line 120, in process
    cropped_pos = self.net_forward(cropped_image)
  File "/home/pakkapon/PRNet/api.py", line 62, in net_forward
    return self.pos_predictor.predict(image)
  File "/home/pakkapon/PRNet/predictor.py", line 98, in predict
    feed_dict = {self.x: image[np.newaxis, :,:,:]})
  File "/home/pakkapon/.virtualenv/prnet/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 950, in run
    run_metadata_ptr)
  File "/home/pakkapon/.virtualenv/prnet/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1149, in _run
    str(subfeed_t.get_shape())))
ValueError: Cannot feed value of shape (1, 256, 256, 1) for Tensor 'Placeholder:0', which has shape '(?, 256, 256, 3)'

Best regrad
Pakkapon

[ikalista]

REALLY helpful!
I was wondering why the number of channels of the image was reduced from 3 to 2 after the rescale()

and your suggestion works