Hi,
Thank you for sharing the project — I’ve been going through it and really appreciate the work you’ve done.
While trying to load the provided .keras model file (MobileNetV2), I ran into an issue. After some debugging, it seems that although the file has a .keras extension, it may have been saved in the older HDF5 format. Since the model includes newer Keras 3.x configurations like groups=1 in DepthwiseConv2D, it throws an error during deserialization — which seems related to the format used during saving.
Would you be open to checking whether the model was saved using the new .keras format? From what I understand, explicitly specifying the format like this should resolve the issue:
model.save("mobilenetv2_skin_classification_model.keras", save_format="keras_v3")
Of course, I could be missing something — so I’d really appreciate your insight if you see it differently.
Thanks again for making the work public — I’ve found it really valuable.
Best regards,
Dan
Hi,
Thank you for sharing the project — I’ve been going through it and really appreciate the work you’ve done.
While trying to load the provided .keras model file (MobileNetV2), I ran into an issue. After some debugging, it seems that although the file has a .keras extension, it may have been saved in the older HDF5 format. Since the model includes newer Keras 3.x configurations like groups=1 in DepthwiseConv2D, it throws an error during deserialization — which seems related to the format used during saving.
Would you be open to checking whether the model was saved using the new .keras format? From what I understand, explicitly specifying the format like this should resolve the issue:
model.save("mobilenetv2_skin_classification_model.keras", save_format="keras_v3")
Of course, I could be missing something — so I’d really appreciate your insight if you see it differently.
Thanks again for making the work public — I’ve found it really valuable.
Best regards,
Dan