CODEWITHSUNDEEP
pythontensorflowkerasconcatenationconv-neural-network
Olivertubs
Olivertubs

Reputation: 45

Multiple Inputs for CNN: images and parameters, how to merge

I use Keras for a CNN and have two types of Inputs: Images of objects, and one or two more parameters describing the object (e.g. weight). How can I train my network with both data sources? Concatenation doesn't seem to work because the inputs have different dimensions. My idea was to concatenate the output of the image analysis and the parameters somehow, before sending it into the dense layers, but I'm not sure how. Or is it possible to merge two classifications in Keras, i.e. classifying the image and the parameter and then merging the classification somehow?

Upvotes: 2

Views: 2570

Answers (1)

Snehal
Snehal

Reputation: 757

You can use Concatenation layer to merge two inputs. Make sure you're converting multiple inputs into same shape; you can do this by adding additional Dense layer to either of your inputs, so that you can get equal length end layers. Use those same shape outputs in Concatenation layer.

Upvotes: 1

Related Questions