Load pretrained keras network using deep learning toolbox

3 views (last 30 days)
nor kadosh
nor kadosh on 10 Aug 2021
Answered: Yongjian Feng on 11 Aug 2021
I'm trying to load pre-trained neural networks by using a deep learning toolbox.
I saved my python model in numbers ways such as:
  • model.save("model")
  • model.save("model.h5")
  • model.save_weights("model.h5")
  • etc.
But none of them working when trying to load them into MATLAB workspace.
model = importKerasLayers("model.h5")
Error using nnet.internal.cnn.keras.ParsedKerasModel (line 26)
Assertion failed.
Error using nnet.internal.cnn.keras.ParsedKerasLayers (line 14)
KM = nnet.internal.cnn.keras.ParsedKerasModel(ModelConfig, Training);
Error in importKerasLayers (line 102)
Layers = nnet.internal.cnn.keras.ParsedKerasLayers(modefilie, varargin{:});

Answers (1)

Yongjian Feng
Yongjian Feng on 11 Aug 2021
Not sure if they are compatible. Why not loading it from python code? Call python from matlab.

Community Treasure Hunt

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

Start Hunting!