CODEWITHSUNDEEP
tensorflowmachine-learningdeep-learningclassificationconvolution
Bastian
Bastian

Reputation: 1593

Tensorflow: How handle multiple inputs

I want to create a convolutional neural network in tensorflow which takes images as as inputs to the first convolution layers and propagates the data from them through the net. At the point where the last pooling layer ist flattened, I want to either add some additional input there or directly to the fully-connected layer.

Note: For each input image of the training data exists a additional set of numerical values which are unique to the image.

Could someone provide some information on how to implement this in tensorflow, please?

Upvotes: 2

Views: 611

Answers (1)

Umang Gupta
Umang Gupta

Reputation: 16440

You can just declare an input variable and use it anywhere you like. It does not have to be at the first layer.

#A,B are placeholders let's say
last_layer = forward(A) # forward function computes to the last layer
output = transform(last_layer,B)

If you add specific code in the question. I can help more.

Upvotes: 1

Related Questions