Style Transfer to Create Real-time Artwork with ML | Fritz

Style Transfer

Use Style Transfer to bring real-time, artistic styles to your apps. Transform photos into masterpieces painted by history’s greatest artists. Great for creative camera apps and user-generated content.
 

Getting Started

import Fritz

let image = FritzVisionImage(buffer: sampleBuffer)

var model: FritzVisionStyleModel?

model.predict(image) { (stylizedBuffer, error) in
    guard let stylizedBuffer = stylizedBuffer else { return }
    self.callbackQueue.async {
        callback(stylizedBuffer)
    }
}
Style Transfer

The Swift code sample here illustrates how simple it can be to use style transfer in your app. Use the links below to access additional documentation, code samples, and tutorials that will help you get started.

Features

Unlimited Styles

We provide 11 styles ready to use. Add your own trained styles to access unlimited creativity.

  • Starry Night by Vincent Van Gogh
  • The Scream by Edvard Munch
  • The Poppy Field by Claude Monet
  • Bicentennial Print from America: The Third Century by Roy Lichtenstein
  • Les Femmes d’Alger by Picasso
  • Head of a Clown by Joseph Kutter
  • Horses on the Seashore by Giorgio de Chirico
  • The Trial by Sidney Nolan
  • Ritmo Plastico by Gino Severini
  • A view through a kaleidoscope
  • Pink and blue rhombuses style
Stable Style Transfer

Visually stabilized style transfer creates beautiful continuity between frames. Available as an upgrade to regular style transfer.

Customize Styles

Our team can create custom style transfer themes upon request.

Or use the Fritz Style Transfer machine learning template to train your own style transfer themes.

Runs On-Device

All predictions / model inferences are made completely on-device.

No internet connection is required to interpret images or video.

No internet dependency means super-fast performance.

Cross-Platform SDKs

Supported mobile platforms:

  • Android Style Transfer
  • iOS Style Transfer
Live Video Performance

Runs on live video with a fast frame rate.

Exact FPS performance varies depending on device, but it is possible to run this feature on live video on modern mobile devices.

Technical Specifications

Architecture

Fast Style Transfer

Model Size

68 KB

17 KB (8-bit quantization)

467 KB (stabilized)

FLOPS

150 M

Input

640x480-pixel image

Arbitrary size images (iOS12)

Output

Stylized image

Formats

Core ML, TensorFlow, TensorFlow Mobile, Keras

Benchmarks

28 FPS on iPhone X

10 FPS on Pixel 2