CONCEPT
Paul Gauguin—a French painter with a distinctive style, renowned for his brushwork and exotic subjects drawn from hot lands. Can a neural network replicate the artist’s unique style and convey the sensations embedded in his paintings? To maintain a contrast between human and machine, the prompts in my project focus on creating icy landscapes rather than replicating the heat of the southern sun, as seen in Gauguin’s work.
SOURCE IMAGES
Fragments of Paul Hogan’s work
CODE
First, we create a notebook in Kaggle, check the settings, and enable the T4 GPU.
Upload the source images and specify their path
setting the style name and its activating prompt
Model Training Settings
RESULT
prompt «an arctic frozen coastline at dawn in PAULGAUGUIN style, cold blue light, snow, ice, expressive brushstrokes, post-impressionist painting, masterpiece»
prompt «a vast arctic tundra with icy hills in PAULGAUGUIN style, cold wind, pale sky, snow-covered ground, vivid painterly texture, post-impressionist painting»
«an arctic seascape with drifting icebergs in PAULGAUGUIN style, deep cold colors, stylized clouds, bold brushwork, post-impressionist masterpiece»
«an arctic winter landscape under a pale sunset in PAULGAUGUIN style, icy mountains, snow fields, dramatic color harmony, post-impressionist painting»
«a remote polar landscape with frozen river and snow plains in PAULGAUGUIN style, cold atmosphere, expressive colors, painterly composition, masterpiece»
The model has successfully learned to mimic the style of a great artist. The original hues of blue, orange, and yellow are preserved in the work. The technique of applying bold brushstrokes and the tendency to round forms have been copied. Potential shortcomings are likely due to the small sample size within the project: only 13 original images and 500 iterations.
Code link:
https://drive.google.com/file/d/18_Pf3lOI0FOgleTvf8QIbBjmCU4bhpqR/view?usp=sharing