Automated Deep Photo Style Transfer
This repository holds a TensorFlow implementation for the paper Automated Deep Photo Style Transfer.
At its core this is a TensorFlow based implementation of the paper Deep Photo Style Transfer.
One of the main contributions of “Automated Deep Photo Style Transfer” is the automatic segmentation of input images and a semantic grouping thereof. Another contribution of this is the optimization of the transfer image by improving the aesthetics of the image with the use of Neural Image Assessment (NIMA).
Examples
Given a content and style image, automatically a segmentation is created and semantically grouped to produce a transfer image in the size of the content image by using the Deep Photo Style Transfer:
Here are some example results (from left to right are the content image, the resulting transfer image and the style image):
Setup
- Download or clone repository files to your computer
- Go into repository folder
- Install requirements:
pip3 install -r requirements.txt --upgrade - Download the weights.zip from the latest release and unzip it into a new folder
weightsunder the project root.
Acknowledgement
- Fujun Luan, Sylvain Paris, Eli Shechtman, Kavita Bala - Deep Photo Style Transfer
- L. Gatys, A. Ecker, M. Bethge - Image Style Transfer Using Convolutional Neural Networks
- A. Levin, D. Lischinski, Y. Weiss - A Closed Form Solution To Natural Image
- H. Zhao, J. Shi, X. Qi, X. Wang, J. Jia - Pyramid Scene Parsing Network
- G. Zhu, C. Iglesias - Sematch: Semantic Entity Search from Knowledge Graph
- Y. Li, Z.A. Bandar, D. Mclean - An approach for measuring semantic similarity between words using multiple information sources
- H. Talebi, P. Milanfar - NIMA: Neural Image Assessment
- H. Talebi, P. Milanfar - Learned Perceptual Image Enhancement
Disclaimer
This software is published for academic and non-commercial use only.