Deep Learning Style Transformations in Ruby

Want to do DeepLearning inference in Ruby?
Yes, you can!

Requirement

• ONNXRuntime and its Ruby binding created by Andrew Kane.
  Liquid error: internal

• Magro and NArray.

• Fast Neural Style Transfer Model in ONNX Model Zoo.

st.rb

require 'magro'
require 'onnxruntime'

model = OnnxRuntime::Model.new('')

input = Magro::IO.imread(ARGV[0])
                 .transpose(2, 0, 1)
                 .expand_dims(0)
                 .to_a

output = model.predict(input1: input)

result = Numo::Int32[*output['output1'][0]]
                    .clip(0, 255)
                    .transpose(1, 2, 0)
                    .cast_to(Numo::UInt8)

Magro::IO.imsave(ARGV[1], result)

Run

Input image size should be a multiple of 4

ruby st.rb in.jpg out.jpg

Generated images

https://commons.wikimedia.org/wiki/File:Arimatsushibori.JPG ¹

candy

mosaic

pointilism

rain_princess

udnie

Beautiful.

Other posts

• YOLO object detection using ONNXRuntime with Ruby
• NArray - Ruby equivalent of NumPy

---

1. Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0)  ↩