There is a new KNIME forum. You can still browse and read content from our old forum but if you want to create new posts or join ongoing discussions, please visit our new KNIME forum:

DL4J FeedForward Learner does not accept picture as input

Member for

4 years 6 months sparel


I am trying to run the workflow published recently in the KNIME Blog (Using DeepLearning4J to classify MNIST digits). After struggling a bit with the Python installation (Protobuf a.o.), I have now the following issue:

The DL4J FeedForward Learner node (Classification) does not seem to recognize/accept a picture column as Feature Column. In the Column Selection Tab, both sides for manual selection are empty, although the pictures are available in the input (and can be visualized in the Image Viewer node).

Am I doing something wrong here?

I'm using KNIME 3.4 with all nodes up-to-date.

Thanks for your help

Thu, 09/28/2017 - 09:53

Member for

2 years 9 months


Hi sparel,

sorry for your troubles. In order to work with images, you additionally need to install the Deeplearning4J - Image Processing extension which can be found in the KNIME Community Contributions Update Site.

I hope that helps.



Thu, 10/05/2017 - 08:06

Member for

4 years 6 months


Thanks, David. It works now.