0
$\begingroup$

I'm new to AI, so bear with me... What would be the easiest way to do this using AI? Where do I start? What technologies are already available?

Sample data: 50.000 before images ex. there's a red ball that needs to be removed 50.000 after images the red ball is removed out of the image (by a human in photoshop)

Feed the data AI learning magic

Feed the AI Input: 1 other before image featuring the red ball Output expected: 1 image where the red ball is removed from the image.

$\endgroup$
0
$\begingroup$

This is a very (very) broad question.

What you are trying to do is utilize some neural model to perform image processing. So your output needs to be the same size as your input, meaning that you need to use a fully-convolutional model for this.

I suggest you read the following article:

Fast Image Processing with Fully-Convolutional Networks

They are doing exactly what you need, only for different image-processing methods (and more than one).

However! I think that before going on to implementing a neural model, you should read a bit about the general subject of machine learning and deep learning. How Do I Learn Neural Networks?

$\endgroup$
  • $\begingroup$ Thank you. I believe it should be a supervised paired image to image translation with a subset of before and after images that are trained in the before and wrongfully edited images added in the lather. Allowing the AI to train on both datasets and cross reference for output. $\endgroup$ – John Doe Jan 1 '19 at 0:28

Not the answer you're looking for? Browse other questions tagged or ask your own question.