Loading saved model from keras importer

21 views (last 30 days)
Khang
Khang on 6 Apr 2023
Edited: Sid on 2 Oct 2025 at 21:28
I tried to importing my model from a different computer using Matlab R022b. And then I got this error:
Error using nnet.internal.cnn.tensorflow.tf2mex
Invalid MEX-file
'C:\ProgramData\MATLAB\SupportPackages\R2022b\toolbox\nnet\supportpackages\keras_importer\+nnet\+internal\+cnn\+tensorflow\tf2mex.mexw64':
The file cannot be accessed by the system.
I used the same Matlab version with the same Deep learning toolbox

Answers (2)

Nihal Reddy
Nihal Reddy on 10 Apr 2023
Edited: Nihal Reddy on 10 Apr 2023
I understand you are getting an error while trying to import model using Keras importer in MATLAB R2022b.
The error message is mostly likely due the model missing a file called 'keras_metadata.pb', it is caused by inappropiate way when saving the model in TensorFlow.
This is because the model is not a Saved Model (https://www.tensorflow.org/guide/saved_model). Please regenerate model and save model using
model.save
instead of
tf.saved_model.save
Thus, the 'keras_metadata.pb' will be included, and MATLAB should import the model successfully.
Currently, "importKerasNetwork"supports TensorFlow-Keras versions as follows:
  • The function fully supports TensorFlow-Keras versions up to 2.2.4.
  • The function offers limited support for TensorFlow-Keras versions 2.2.5 to 2.4.0.
Refer to the following documentation link for more information and examples-

Sid
Sid on 2 Oct 2025 at 19:08
Edited: Sid on 2 Oct 2025 at 21:28
Hi, just need to revive this thread, since I am facing this exact problem while trying out the example:
Brain MRI Segmentation Using Pretrained 3-D U-Net Network
Early in the example, it instructs me to use :
net = importNetworkFromTensorFlow(fullfile(dataDir,"trainedSynthSegModel")).
Initially, I installed the module since it was not available, but after imstallation,
subsequent runs gave me the following error message, pasted below. running dependency walker on the tf2mex.mexw64 shows many missing dlls, almost all of them are libmex* or similar:
-------------------------
LIBMATLABDATAARRAY.DLL
LIBMEX.DLL
LIBMWFL.DLL
LIBMWFOUNDATION_FILESYSTEM.DLL
LIBMWI18N.DLL
LIBMWSPROOTUTILS.DLL
LIBMX.DLL
LIBUT.DLL
---------------------------------------
How should I proceed from this point onwards? I tried changing/adding the PATH variables in system settings, but I am still getting this error. executing !path from matlab command window whows me the path where ther libs are.
Hope to get some constructive help on this!Thanks!
Sid.
Invalid MEX-file 'C:\ProgramData\MATLAB\SupportPackages\R2025a\toolbox\nnet\supportpackages\keras_importer\+nnet\+internal\+cnn\+tensorflow\tf2mex.mexw64': The specified module could not be found.
Error in
smstruct = jsondecode(nnet.internal.cnn.tensorflow.tf2mex('saved_model', fullfile(obj.SavedModelPath, 'saved_model.pb')));
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error in
sm = savedmodel.TFSavedModel(this.ModelPath, this, true);
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error in
net = importManager.translate;
^^^^^^^^^^^^^^^^^^^^^^^
Error in
Network = nnet.internal.cnn.tensorflow.importTensorFlowNetwork(modelFolder, 'TargetNetwork', 'dlnetwork', 'PackageName', options.PackageName, 'Namespace',options.Namespace, 'OnlySupportDlnetwork', true, 'PreferredNestingType', options.PreferredNestingType);
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Error in
Network = nnet.internal.cnn.tensorflow.importNetworkFromTensorFlow(modelFolder, varargin{:});
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Categories

Find more on Image Data Workflows in Help Center and File Exchange

Products


Release

R2022b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!