Search code examples
pythontensorflowdeep-learningtflearn

Generate data with neural networks

I want to generate data with a neural network which I have trained before. Did someone know how to make that with TensorFlow of TFlearn?

It should generate a image with a network i trained before with a dataset contains pictures.

Solution

  • I don't know how you imagined "generate an image", but one way would be trying to see what image maximizes your prediction for one of your labels. To do that you can replace you input image placeholder by a variable you initialize with random noise. Your optimizer will optimize this variable to give a maximum prediction.

    An example is given in lesson 4 of the Creative Applications Of Deep Learning course on Kadenze. Check it out here: https://github.com/pkmital/CADL/blob/master/session-4/lecture-4.ipynb