fast-style-transfer

Fast Style Transfer

Add styles from famous paintings to any photo! This work is just an example demonstration from lengstrom/fast-style-transfer amazing work

Requirements

Run pip install -r requirements.txt

This will install numpy, opencv and tensorflow, which are it's only dependencies.

How to run

Use PainterNetwork class from painter.py.

The class can be used as shown in the following example:

from painter import PainterNetwork
import cv2
import sys
import numpy as np


# Pick one of the 6 available style models
MODEL_PATH = "/path/to/style/model"

IMG_PATH = "/path/to/image"

# Read image into [0-255] numpy array
img = cv2.imread(IMG_PATH)

# Remember openCV works in BGR
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# `run` method receives [0-255] RGB image array and model path
# returns [0-255] RGB image array.
img = PainterNetwork.run(img, MODEL_PATH) # Receives 

# Remember openCV works in BGR
img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)

# Display image
cv2.imshow('image', img)
cv2.waitKey(0)
cv2.destroyAllWindows()

# Save image
cv2.imwrite("output.png", img)

How to use other styles?

You will need to go to the original repo lengstrom/fast-style-transfer to create new networks with your own styles. Once you got your new .ckpt file, you can come back to these snippets to run it.

Model info

Models were found in lengstrom/fast-style-transfer.