Awesome
Deep Feature Rotation for Multimodal Image Style Transfer
This repository contains the official implementation of paper: <br> Deep Feature Rotation for Multimodal Image Style Transfer <br> Son Truong Nguyen, Nguyen Quang Tuyen, Nguyen Hong Phuc <br> In NICS 2021 Oral.<br>
Paper (arXiv version) | Paper (IEEE version) | Presentation | Colab Demo | Bibtex |
---|
Table of Content
Overview
We propose a simple method for representing style features in many ways called Deep Feature Rotation (DFR), while still achieving effective stylization compared to more complex methods in style transfer. Our approach is a representative of the many ways of augmentation for intermediate feature embedding without consuming too much computational expense.
<br> <p align="center"> <img src="doc/model.png" alt="Model architecture" width="800"> </p> <p align="center"> <sub><em>Model architecture.</em></sub> </p>Getting started
Demo
<p align="center"> <a href="https://colab.research.google.com/drive/1nmf4_YnUBq5dGGTgWeN1fYNYOSOKeQ-1?usp=sharing"> <img src="https://colab.research.google.com/assets/colab-badge.svg"/> </a> <br> Try out in Google Colab </p>Installation
-
Clone this repository and check the
requirements.txt
:git clone https://github.com/sonnguyen129/deep-feature-rotation cd deep-feature-rotation pip install -r requirements.txt
-
Inference:
- Prepare your content image and style image. I provide some in the
data/content
anddata/style
and you can try to use them easily. - Simply run:
python train.py --content-path <CONTENT_PATH> --style-path <STYLE_PATH>
The test results will be saved to
./results
by default. - Prepare your content image and style image. I provide some in the
Results
<p align="center"> <img src="doc/rotation_weight.png"> </p> <p align="center"> <sub><em>Experimental result in different rotation weight.</em></sub> </p> <br>Extensive results
We provide a visual comparison between other rotation angles that do not appear in the paper. The rotation angles will produce a very diverse number of outputs. This has proven the effectiveness of our method with other methods.
Citation
If you find this work useful for your research, please cite:
@INPROCEEDINGS{9701465,
author={Nguyen, Son Truong and Tuyen, Nguyen Quang and Phuc, Nguyen Hong},
booktitle={2021 8th NAFOSTED Conference on Information and Computer Science (NICS)},
title={Deep Feature Rotation for Multimodal Image Style Transfer},
year={2021},
pages={260-265},
doi={10.1109/NICS54270.2021.9701465}
}
Contact
If you have any questions/comments/bug reports, feel free to open a github issue or pull a request or mail to the author Son Truong Nguyen.