Come upon a network with the ability to transfer content from paintings to sketches
Closed this issue · 6 comments
The algorithm of neural style-transfer has the ability to combine the content image and style image, but it does not work for sketches and paintings. I came upon an algorithm that can do similar things on sketches.
It should be taken into consideration that the result is predicted only after about 8,000 iterations.
However I only has about 4,000 images to train my model and I am very interested in whether my algorithm can reach the state of art. So could you share some of training data to help me?
Hi , I also tried some network to use some style hint info.
If you can be our part-time employee, we can share our dataset , but we couldn't share it openly.
If you have some interest, please send apply mail https://www.preferred-networks.jp/en/job
After feed my network with about 10,000 pictures, I randomly selected some results.
| Sketch | Style | Result |
|---|---|---|
 |
 |
 |
|
 |
 |
|
 |
 |
 |
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
|
 |
 |
I am still very interested in your very big training data set. I would like to talk with you in slack. May I take a moment of your time and have a brief talk?
oh looks much better.
Selecting the coloring style is very interesting topic, and I want to implement to the service too.
Yes, ofcourse
Very interesting! Thanks for sharing!!