When I try to load a keras model using importKerasNetwork I get the following error (R2020A)

When I try to load a keras model using importKerasNetwork I get the following error (this is in R2020A).

Brace indexing is not supported for variables of this type.

Error in nnet.internal.cnn.keras.readModelAndTrainingConfigs>iVerifyKerasBackend

(line 71)

if ~isequal(BackendCell{1}, 'tensorflow')

Error in nnet.internal.cnn.keras.readModelAndTrainingConfigs (line 8)

iVerifyKerasBackend(ConfigFile);

Error in nnet.internal.cnn.keras.importKerasNetwork (line 21)

[ModelConfig, TrainingConfig, KerasVersion] =

nnet.internal.cnn.keras.readModelAndTrainingConfigs(ConfigFile, Format);

Error in importKerasNetwork (line 91)

Network = nnet.internal.cnn.keras.importKerasNetwork(modelfile, varargin{:});

Error in LoadingEyetoneIntoMatlab (line 4)

net=importKerasNetwork(sprintf('%s%s', modelpath,networkname));

Answer 1

[Sourabh Kondapaka]

Hi Alexander Berardino,

In the below line:

net=importKerasNetwork(sprintf('%s%s', modelpath,networkname));

Between the '%s' and '%s' , you'll need to add a '/' .

Although this is not an ideal approach, you can do something like this:

modelPath = fileparts(which(modelName));
net = importKerasnetwork([modelPath filesep modelName '.h5']); %Assuming your model has .h5 extension

To learn more about the functions used above, refer the following links:

which

fileparts

filesep

[Alexander Berardino]

Hi Sourabh,

Thanks for your response. The error is not actually in the file path (the pathname variables in the error I copied above have '/' in the correct place.)

I tried your version, and I get the same error. It appears there is an error in nnet.internal.cnn.keras.readModelAndTrainingConfigs>iVerifyKerasBackend here if ~isequal(BackendCell{1}, 'tensorflow'), but when inspecting BackendCell, it appears that BackendCell is not a cell array, but rather a character array that says 'tensorflow'.

It seems as though there is a mismatch between the way that character array is being loaded in and what the matlab function is expecting.