CODEWITHSUNDEEP
kerasdeep-learningocrcoremltext-recognition
sariii
sariii

Reputation: 2140

OCR code written without custom loss function

I am working on OCR model. my final goal is to convert OCR code into coreML and deploy it into ios.

I have looked and run a couple of the github source codes namely:

here

here

as you have a look on them they all implemented loss as a custom layer with lambda layer.

the problem start when I want to convert this to coreML.

my piece of the code to convert to CoreMl:

import coremltools

def convert_lambda(layer):
    # Only convert this Lambda layer if it is for our swish function.
    if layer.function == ctc_lambda_func:
        params = NeuralNetwork_pb2.CustomLayerParams()

        # The name of the Swift or Obj-C class that implements this layer.
        params.className = "x"

        # The desciption is shown in Xcode's mlmodel viewer.
        params.description = "A fancy new loss"

        return params
    else:
        return None


print("\nConverting the model:")

# Convert the model to Core ML.
coreml_model = coremltools.converters.keras.convert(
    model,
    # 'weightswithoutstnlrchangedbackend.best.hdf5',
    input_names="image",
    image_input_names="image",
    output_names="output",
    add_custom_layers=True,
    custom_conversion_functions={"Lambda": convert_lambda},
    )

but it raises error 

Converting the model:
Traceback (most recent call last):
  File "/home/sgnbx/Downloads/projects/CRNN-with-STN-master/CRNN_with_STN.py", line 201, in <module>
    custom_conversion_functions={"Lambda": convert_lambda},
  File "/home/sgnbx/anaconda3/envs/tf_gpu/lib/python3.6/site-packages/coremltools/converters/keras/_keras_converter.py", line 760, in convert
    custom_conversion_functions=custom_conversion_functions)
  File "/home/sgnbx/anaconda3/envs/tf_gpu/lib/python3.6/site-packages/coremltools/converters/keras/_keras_converter.py", line 556, in convertToSpec
    custom_objects=custom_objects)
  File "/home/sgnbx/anaconda3/envs/tf_gpu/lib/python3.6/site-packages/coremltools/converters/keras/_keras2_converter.py", line 255, in _convert
    if input_names[idx] in input_name_shape_dict:
IndexError: list index out of range
Input name length mismatch

I am kind of not sure I can resolve this as I did not find anything relevant to this error to resolve.

In other hand most codes for OCR have Custom Loss function which probably again I face with the same problem.

So in the end I have two question:

  1. Do you know how to resolve this error
  2. my main question do you know any source code for OCR which is in KERAS (As i have to convert it to coreMl) and do not have custom loss function so it will be ok converting to CoreMl without problem?

Thanks in advance:)

just to make my question thorough:

this is the custom loss function in the source I am working:

    def ctc_lambda_func(args):
        iy_pred, ilabels, iinput_length, ilabel_length = args
        # the 2 is critical here since the first couple outputs of the RNN
        # tend to be garbage:
        iy_pred = iy_pred[:, 2:, :]  # no such influence
        return backend.ctc_batch_cost(ilabels, iy_pred, iinput_length, ilabel_length)
    loss_out = Lambda(ctc_lambda_func, output_shape=(1,), name='ctc') 
([fc_2, labels, input_length, label_length])

and then use it in compile:

model.compile(loss={'ctc': lambda y_true, y_pred: y_pred}, optimizer=sgd)

Upvotes: 0

Views: 189

Answers (1)

baaj2109
baaj2109

Reputation: 11

CoreML doesn't allow you train model, so it's not important to have a loss function or not. If you only want to use CRNN as predictor on iOS , you should just convert base_model in second link.

Upvotes: 1

Related Questions