TypeError: 'InputLayer' object is not iterable with CoreMLTools

问题

I'm trying to convert a VGG model to coremltools. When I run the following code to convert the model:

with CustomObjectScope({'relu6': keras.layers.ReLU,'DepthwiseConv2D': keras.layers.DepthwiseConv2D}):
    from keras.models import load_model
    import coremltools


model_directory = 'KerasModels/VGG-7-3-20_13categories.h5'

keras_model = load_model(model_directory)
input_layer = InputLayer(input_shape=(224, 224, 3), name="input_1")

# Save and convert :
keras_model.layers[0] = input_layer
keras_model.save(model_directory)
print("Changed2")


your_model = coremltools.converters.keras.convert(model_directory, input_names=['image'], output_names=['output'], image_input_names='image')




your_model.save('RecycleNet.mlmodel')

I get the following error:

TypeError: 'InputLayer' object is not iterable

How should I go about converting this model to coremltools? Thanks

回答1:

I fixed this error by switching from using:

coremltools.converters.keras

to:

coremltools.converters.tensorflow

回答2:

This line of code solved for me:

coremlModel = coremltools.convert(model)

instead of using this:

coremlModel = coremltools.converters.keras.convert(model)

来源：https://stackoverflow.com/questions/62719352/typeerror-inputlayer-object-is-not-iterable-with-coremltools