Anyone missed this great plug-in in GIMP 2.10.
Well I tried to run it in 2.10.8 and it failed so I found out I only had to change the tile calculations from 64 to 128 and it works flawlessly in GIMP 2.10.8
More details on plug-in and download here --> https://gimplearn.net/viewtopic.php?f=3&t=2357

Is there a way to have it as non .py?
I am having problems with the .py files.

If you have problem with .py files you should look into fixing that.

GREAT, Tin !
Thanks a lot.

For me it works very well - on the condition that the sizes Source and Target match.
Thank you very much for reminding this interesting plugin.

For me it works also using images of different sizes.
Target (the image which should have colours changed):

Source (the image from which colours should be taken):


Outcome from the filter:


Thanks again a lot, Tin.

(both images from Deviant Art)

Thanks !!! for the plugin

I like it.


Doron.

Thank you! This was very good plug-in, glad to see it resurface.

I THOUGHT some people might miss this filter.
Thank you Dinasset for the great filter.
It was a total fluke that I found the 128 instead of 64 error. If it showed some other error I might given up easier.

Thanks for the fix. This filter is great for transferring correct colors to similar photos. By the way i had no problems with 2 different size images either.

The reason for my mistake was to open (source and target) as a layers - instead of separate images.

Interesting...

Thanks Dinasset and Tim for this. I find color transfer interesting after recently experimenting with style transfer. I got 'neural style transfer' working on this box after installing Tensorflow, CUDA and more. After getting neural style transfer running I can convert any photo into an impressionist painting. And thats it really. I don't even like impressionist painting that much but for those who do, remember machines can now do it as well as humans

What steps are all involved in getting neural styles to work on your machine I'd like to try it see if it's any better than submitting to a site.
And also doesn't the style change depending on your style image input? or does it only do impressionist?

Thanks Diego and Tin

A couple of hard to answer questions.
For Windows try here
This is interesting
I used this
"does it only do impressionist?"
The finished works are produced entirely by machines (codes, algorithms etc). At the moment, machines are not seeing the world the same way humans do.





This is what I got using a Dali painting as the style image. I don't think it's anything to get excited about. Unless you have a spare machine and plenty of spare time I don't recommend neural style transfer. (use the online upload sites)
There is plenty of Python code to play with and you are probably smarted than me so you may get better results. LOL.
It's just a matter of time before we start buying art produced by machines.

thanks, this will be very useful plugin for my future gimp art

Tas_mania, Thanks for the info. I'll look into the one you used as that one seems to have lots of options to play with like the masking for multiple styles and option to transfer only luminance preserving colors.

UPDATE: Also when you upload to sites for neural styles, they have an image size limit (usually small). I am running the neural style you used on my machine right now (no CUDA using cpu so it might be slow as hell) and it seems to be running fine on large image. hopefully output is same size as content image. If that's the case then it's better than uploading to sites because of no size limit (but I cross my fingers, because I havn't seen any output yet just happy to see it running).

Lylejk might get into this because I remember vaguely that he complainted about the image size limit being a show stopper.

3 hours and still no result :O