A command-line tool and Python library to automatically generate new textures similar to a source image or photograph. It's useful in the context of computer graphics if you want to make variations on a theme or expand the size of an existing texture.

This tool is powered by deep learning technology — using a combination of convolution networks and example-based optimization to synthesize images. We're aiming to make neural-texturize the highest-quality open source library available!

1. Examples & Demos
2. Installation
3. Commands & Usage

The examples are available as notebooks, and you can run them directly in-browser thanks to Jupyter and Google Colab:

• Gravel — online demo and source notebook.
• Grass — online demo and source notebook.

These demo materials are released under the Creative Commons BY-NC-SA license, including the text, images and code.

If you're a developer and want to install the library locally, start by cloning the repository to your local disk:

git clone https://github.com/photogeniq/neural-texturize.git

Then, you can create a new virtual environment called myenv by installing Miniconda and calling the following commands, depending whether you want to run on CPU or GPU (via CUDA):

cd neural-texturize # a) Use this if you have an *Nvidia GPU only*. conda env create -n myenv -f tasks/setup-cuda.yml # b) Fallback if you just want to run on CPU. conda env create -n myenv -f tasks/setup-cpu.yml

Once the virtual environment is created, you can activate it and finish the setup of neural-texturize with these commands:

conda activate myenv poetry install

Finally, you can check if everything worked by calling the script:

texturize

You can use conda env remove -n myenv to delete the virtual environment once you are done.

The main script takes a source image as a texture, and generates a new output that captures the style of the original. Here are some examples:

texturize samples/grass.webp --size=1440x960 --output=result.png texturize samples/gravel.png --iterations=200 --precision=1e-5 texturize samples/sand.tiff --output=tmp/{source}-{octave}.webp texturize samples/brick.jpg --device=cpu

For details about the command-line options, see the tool itself:

texturize --help

Here are the command-line options currently available:

Usage: texturize SOURCE... [--size=WxH] [--output=FILE] [--variations=V] [--seed=SEED] [--mode=MODE] [--octaves=O] [--threshold=H] [--iterations=I] [--device=DEVICE] [--precision=PRECISION] [--quiet] [--verbose] Options: SOURCE Path to source image to use as texture. -s WxH, --size=WxH Output resolution as WIDTHxHEIGHT. [default: 640x480] -o FILE, --output=FILE Filename for saving the result, includes format variables. [default: {source}_gen{variation}.png] --variations=V Number of images to generate at same time. [default: 1] --seed=SEED Configure the random number generation. --mode=MODE Either "patch" or "gram" to specify critics. [default: gram] --octaves=O Number of octaves to process. [default: 5] --threshold=T Quality for optimization, lower is better. [default: 1e-4] --iterations=I Maximum number of iterations each octave. [default: 99] --device=DEVICE Hardware to use, either "cpu" or "cuda". --precision=PRECISION Floating-point format to use, "float16" or "float32". --quiet Suppress any messages going to stdout. --verbose Display more information on stdout. -h, --help Show this message.