Create a Custom Image Inpainter

Inpaint is a generic function to fill in missing values in an image. If the image domain is specific, one can obtain better results with a custom image inpainter. This example shows how to create such a custom inpainter using LearnDistribution and SynthesizeMissingValues.

Obtain a dataset of images.

Train a distribution on the pixel values of the images.

Fill in missing values (defined by gray pixels) of new images.

Compare with the results obtained by the generic Inpaint function.