Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/devkihyun/neural-style-transfer-tensorflow-keras
Tensorflow(using keras pretrained model) implementation of 'Image Style Transfer Using Convolutional Neural Networks'
https://github.com/devkihyun/neural-style-transfer-tensorflow-keras
deep-neural-networks keras python style-transfer tensorflow
Last synced: 4 days ago
JSON representation
Tensorflow(using keras pretrained model) implementation of 'Image Style Transfer Using Convolutional Neural Networks'
- Host: GitHub
- URL: https://github.com/devkihyun/neural-style-transfer-tensorflow-keras
- Owner: DevKiHyun
- Created: 2019-05-21T06:34:14.000Z (over 5 years ago)
- Default Branch: master
- Last Pushed: 2019-07-09T12:05:29.000Z (over 5 years ago)
- Last Synced: 2024-07-25T11:10:36.359Z (5 months ago)
- Topics: deep-neural-networks, keras, python, style-transfer, tensorflow
- Language: Jupyter Notebook
- Homepage:
- Size: 5.06 MB
- Stars: 6
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Neural Style Transfer-TF&Keras (2019/05/20)
## Introduction
I implement a tensorflow&keras model described in the papers
- ["A Neural Algorithm of Artistic Style"](https://arxiv.org/pdf/1508.06576v2.pdf).
- ["Image Style Transfer Using Convolutional Neural Networks", CVPR 2016](https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Gatys_Image_Style_Transfer_CVPR_2016_paper.pdf). by Leon A. Gatys, Alexander S. Ecker, Matthias Bethge
## Environment
- Ubuntu 16.04
- Python 3.6
## Depenency
- Numpy
- Opencv2
- Matplotlib
- Tensorflow (1.4 <= x <=1.13)
## Files
- `images(dir)` : images for style transfer.
- `neural_style_transfer.py` : main code for style transfer.
- `neural_style_transfer.ipynb` : you can execute this code on jupyternotebook. recommend 'Colab' of Google.
## How to use
#### Running
```
python neural_style_transfer.py --content_image_path --style_image_path
# ex)
python neural_style_transfer.py --content_image_path images/content/tubingen.jpg --style_image_path images/style/starry-night.jpg
```
#### Arguments
*Required* :
- `--content_image_path`: Path of the content image. *Default*: `images/content/tubingen.jpg`
- `--style_image_path`: Path of the style image. *Default*: `images/style/starry-night.jpg`
*Optional* :
- `--model_type`: Type of pretrained model. *Choices*: 0(VGG16), 1(VGG19) *Default*: `0`
- `--image_resize`: Resize(int or tuple) images. *Default*: `512`
- `--rescale_image`: Rescale the final image to original size. *Default*: `False`
- `--content_blocks`: Layer list for feature vector of Content image. *Default*: `['block4_conv2']`
- `--style_blocks`: Layer list for feature vector of Style image. *Default*: `['block1_conv1', 'block2_conv1', 'block3_conv1', 'block4_conv1', 'block5_conv1']`
- `--loss_ratio`: Weight of content-loss relative to style-loss. Alpha over beta in the paper. *Default*: `1e-3`
- `--total_variation_weight`: Total Variation weight. *Default*: `0 (If you want to use this, then 8.5e-5)`
- `--initial_type`: The initial image to generate image. *Choices*: content, style, random. *Default*: `'random'`
- `--optimizer_type`: The optimizer for optimization. *Choices*: 0(Adam Optimizer), 1(L-BFGS-B Optimizer). *Default*: `1`
- `--learning_rate`: The value of learning rate for Adam Optimizer. *Default*: `1e+1`
- `--beta_1`: Beta_1 of Adam Optimizer. *Default*: `0.9`
- `--beta_2`: Beta_2 of Adam Optimizer. *Default*: `0.999`
- `--epsilon`: Epsilon of Adam Optimizer. *Default*: `1e-08`
- `--iteration`: The number of iterations. *Default*: `150`
## Sample results
### Content Image : Tübingen, Germany
### Style Image : starry-night, seated-nude, shipwreck, kandinsky, the_scream
All of results made from default setting and executed on Colab.
<p align="center">
<img src="images/content/tubingen.jpg" height="192px">
<img src="images/sample/tubingen_shipwreck.jpg" height="192px">
<img src="images/sample/tubingen_starry_night.jpg" height="192px">
<img src="images/sample/tubingen_seated_nude.jpg" height="192px">
<img src="images/sample/tubingen_the_scream.jpg" height="192px">
<img src="images/sample/tubingen_kandinsky.jpg" height="192px">
</p>
### Content Image :
#### Female Knight ([source](https://www.artstation.com/artwork/4zXxW))
#### blue-moon-lake ([source](https://github.com/titu1994/Neural-Style-Transfer))
### Style Image : wave
All of results made from default setting and executed on Colab.
</p>
<p align='center'>
<img src = 'images/content/blue-moon-lake.jpg' height = '210px'>
<img src = 'images/style/wave.jpg' height = '210px'>
<br>
<img src = 'images/sample/blue_moon_lake_wave.jpg' '365px' width = '710px'>
<br>
<img src = 'images/content/female_knight.jpg' height = '210px'>
<img src = 'images/style/wave.jpg' height = '210px'>
<br>
<img src = 'images/sample/female_knight_wave.png' height = '365px' width = '710px'>
</p>
## Reference
#### https://github.com/anishathalye/neural-style.
#### https://github.com/titu1994/Neural-Style-Transfer
#### https://github.com/hwalsuklee/tensorflow-style-transfer
#### https://github.com/cysmith/neural-style-tf