Python script to quickly generate a Font Awesome icon imposed on a background, mostly intended to be used to steer the usage of a VQGAN + CLIP model generation, such as with this Colab Notebook.

Inspired by Mark Ridel's attempt at controlling figure placement for VQGAN + CLIP.

You can the dependences via:

pip3 install Pillow numpy fire icon_font_to_png

Download this repo from GitHub and cd into the downloaded folder.

The main usage is a icon_image.py script, which can be used to generate an icon image.

For example, with the default usage:

python3 icon_image.py

Generates this image:

If you want to generate an icon in the shape of the Facebook logo with appropriate colors and no background noise, you could do:

python3 icon_image.py --icon_name "fab fa-facebook-f" --bg_color white --icon_color "#4267B2" --bg_noise False

If you want to create a rectangular image with an icon aligned right, you can do:

python3 icon_image.py --icon_name "fas fa-male" --bg_color "#34ebeb" --bg_width 480 --bg_height 360 --icon_size 300 --align "right"

• icon_name: The Font Awesome icon name, including the fa- prefix. (e.g. fab fa-apple)
• sketch_path: Takes a sketch on a pure white background as input and composites it onto the noise, with the white background made transparent. If specified, this behavior replaces the icon generation.
• bg_color: Background color, in text or hex.
• icon_color: Icon color, in text or hex.
• bg_noise: Whether to add uniform noise to the background. (default: True)
• bg_noise_opacity: The strength of the noise, in range [0, 1]. (default: 0.2)
• rainbow_noise: Whether the noise should be rainbow, vs. black-and-white. If enabled, you should increase bg_noise_opacity (default: False)
• icon_opacity: The opacity of the icon from [0, 1]. (default: 1.0 / no transparency)
• bg_width: Width of the final image (default: 600)
• bg_height: Height of the final image (default: 600)
• icon_size: Size of the icon within the final image (and the size the base icon is rendered at) (default: 500)
• icon_width: Width of the icon in the final image; uses icon_size if not specified (default: None)
• icon_height: Height of the icon in the final image; uses icon_size if not specified (default: None)
• seed: Seed for random noise generation. (default: 42)
• align: Alignment for the icon within the final image. (default: center: valid values are center, left, right, top, bottom)

• The default parameters of 600px for the image and 500px for the icon are chosen to be the largest compatible with the default parameters of the VQGAN + CLIP Colab Notebook without going OOM.
• When using it as a initial/target in the VQGAN + CLIP Colab Notebook, it's help to set background and icon colors to be similar to what you want in the final result.
• The use of the background noise theoetically helps VQGAN + CLIP differentiate between the icon and the background, but experimentation will always be useful.
• The default icon color, #7b7568, roughly corresponds to the RGB color of the mean ImageNet image, which the VQGAN + CLIP Colab Notebook normalizes.

Max Woolf (@minimaxir)

Max's open-source projects are supported by his Patreon and GitHub Sponsors. If you found this project helpful, any monetary contributions to the Patreon are appreciated and will be put to good creative use.

MIT

Font Awesome icon font files included per the terms in its SIL OFL 1.1 License.