Art Transfer AI is full of different algorithms that are able to transfer the art style of one image (ex: Van Gogh picture) to other image, obtaining a mixture between two images. You can draw your picture in any artistic style you want!

After the "Convolutional Neural Networks" course at Coursera (Deeplearning.ai specialization), I created mDL-ArtTransfer: https://github.com/muntisa/mDL-ArtTransfer.

Thus, I compared the script from Week4: Face recognition and Neural style transfer, practice Neural Style Transfer with other three algorithms for the same task. Therefore, this git is a benchmark of AI Art Transfer results usinng default parameters using four sources:

• Coursera : https://www.coursera.org/learn/convolutional-neural-networks
• fchollet : https://github.com/keras-team/keras/blob/master/examples/neural_style_transfer.py
• anishathalye : https://github.com/anishathalye/neural-style
• ShafeenTejani: https://github.com/ShafeenTejani/style-transfer

Hare is an example of mixing a dog with my painting style using four different AI scripts:

One of the problem was the transparency in the PNGs files. Therefore, I created the second repo based on fchollet 's AI - DL-Art4Transparency: https://github.com/muntisa/DL-Art4Transparency.

Below you can see an example where a leaf image with transparency is mixed with the Van Gogh's style (Starry Night). You can observe that the transparency for all pixels is maintained, even the shining of the leaf!

In the future, I will place more example such as the modification of your game assets with a specific style ;-)

Have a nice AI experience!
muntisa