I just tried to run all workflows you mentioned about Keras and figured out that all are running very well.
For me, it must be a very good starting point to delve into deep learning using KNIME platform.
If you allow me to ask more, some sophisticated model like LSTM, GAN, etc will be a good asset in terms of DL model
and some application workflows in conjunction with other KNIME nodes like object detections & countings, regression for industrial use, text mining etc would be another benifit to this open innovative society.
Anyway, I am more than happier to know KNIME has the very strong functionality and I thank you so much to all in KNIME.