I've converted model from caffe and trying to run predict like this:
e = np.zeros((1,3,224,224))
d = {}
d['data'] = e
r = coreml_model.predict(d)And have
RuntimeError: {
NSLocalizedDescription = "The model expects input feature data to be an image, but the input is of type 5.";
}
Any ideas?
Thanks!
For Images, CoreML uses python's Pillow library (pip install Pillow). Here is the code snippet that should work for you
import coremltools
# Load an image using PIL
from PIL import Image
rose = Image.open('rose.jpg')
coreml_model.predict({'data': rose})An example of doing this is also availiable in Session 710's video (https://developer.apple.com/videos/play/wwdc2017/710/)
Note that in your code above, you were passing a Numpy object which gets converted to an MLMultiArray in CoreML.
It works, thanks!
I'm trying to do the same, following the WWDC video, but I'm receiving an error:
Error: value type not convertible:
<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=200x200 at 0x121AC04D0>
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-38-82fd85392a7f> in <module>()
----> 1 model.predict({'stampImage': image})
/anaconda/envs/coremltools_py2/lib/python2.7/site-packages/coremltools/models/model.pyc in predict(self, data, **kwargs)
236 """
237 if self.__proxy__:
--> 238 return self.__proxy__.predict(data)
239 else:
240 if _sys.platform != 'darwin' or float('.'.join(_platform.mac_ver()[0].split('.')[:2])) < 10.13:
RuntimeError: value type not convertibleThe Keras model works fine... Do you have any idea about what is going wrong?
e = np.zeros((3,224,224))
d = {}
d['data'] = e
r = coreml_model.predict(d)The model expects the input to be of size 1, 3 or 5. So just change (1,3,224,224) which is of size 4 to (3,224,224) which is of size 3.