gpkg.magenta.image 0.3.0.dev2

Image generators from the Magenta project

Models

magenta-arbitrary-stylize

Fast artistic style transfer using arbitrary painting styles

Operations

generate

Generate a stylized using model pretrained on PNB and DRD images

Flags

content-images

Path to content images (include glob pattern matching images) (required)

image-size

Size of images (default is 1024)

interpolation-weights

Interpolation weights This value is a list of float values inside square brackets. Each value is a weight for interpolation between the parameters of the identity transform and the style parameters of the style image. The larger the weight is the strength of stylization is more. Weight of 1.0 means the normal style transfer and weight of 0.0 means identity transform. (default is ‘[1.0]’)

style-images

Path to style images (include glob pattern matching images) (required)

References

• https://github.com/tensorflow/magenta/tree/master/magenta/models/arbitrary_image_stylization

• https://arxiv.org/abs/1705.06830

• https://arxiv.org/abs/1610.07629

• https://arxiv.org/abs/1603.08155

• https://arxiv.org/abs/1508.06576

magenta-image-stylize

Implementation of ‘A Learned Representation for Artistic Style’

Operations

generate

Generate a stylized image using a pretrained models

Flags

image

Image to stylize (required)

style

Style to apply (monet or varied) (required)

References

• https://github.com/tensorflow/magenta/tree/master/magenta/models/image_stylization

• https://arxiv.org/abs/1610.07629