Saving apply_transform parameters to training data

[rgodb...@gmail.com]

The gist of what I am trying to do:

I am trying to take create a neural network that will be able to detect the angle of rotation and zoom of an image. I am trying to create the training data by using the apply_transform function of ImageDataGenerator by first zooming on the image up to four times, each time rotating the image 60 times. I can make the images no problem, but how do I save the parameters of the transformation to a structure that can be used as training data?

[Lance Norskog]

The altered image is the input, and the alteration coefficients are the label.

I would try writing a custom Generator class that applies the transforms and fills in the label data.

On Mon, Dec 2, 2019 at 7:18 PM <rgodb...@gmail.com> wrote:

The gist of what I am trying to do:

I am trying to take create a neural network that will be able to detect the angle of rotation and zoom of an image. I am trying to create the training data by using the apply_transform function of ImageDataGenerator by first zooming on the image up to four times, each time rotating the image 60 times. I can make the images no problem, but how do I save the parameters of the transformation to a structure that can be used as training data?

--
You received this message because you are subscribed to the Google Groups "Keras-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to keras-users...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/keras-users/202fe218-9d9e-4f8c-b7ff-8e3dbeb8596b%40googlegroups.com.

--

Lance Norskog
lance....@gmail.com
Redwood City, CA