Style transfer for augmented city

Project idea

Using fast-style-transfer neural networks techniques we generate and transfer images from cities, using images from forest and other green and blue areas.

The project is one of the projects of City Interaction lab in Paris https://sites.google.com/cri-paris.org/cityinteractionlab/

Technical details

Using the package developed by Gene Kogan https://github.com/genekogan/fast-style-transfer and inspired from @lengstrom (GitHub repository https://github.com/lengstrom/fast-style-transfer) we first follow the installation patterns as suggested in fast-style-transfer:

Step 1：Install Anaconda

https://docs.anaconda.com/anaconda/install/

Step 2：Build a virtual environment

Run the following commands in sequence in Anaconda Prompt:

conda create -n tf-gpu tensorflow-gpu=2.1.0
conda activate tf-gpu
conda install jupyterlab
jupyter lab

Run the following command in the notebook or just conda install the package:

!pip install moviepy==1.0.2

Then we follow the commands below to use fast-style-transfer:

Before the next lines you should run setup.sh from https://github.com/lengstrom/fast-style-transfer

Then follow:
python style.py --style path/to/style/img.jpg \
--checkpoint-dir checkpoint/path \
--test path/to/test/img.jpg \
--test-dir path/to/test/dir \
--content-weight 1.5e1 \
--checkpoint-iterations 1000 \
--batch-size 20

What style.py is doing? style.py trains networks that can transfer styles from artwork into images.

We need to create checkpoint-directory and then call the example of code:

python style.py --style example/style/la_muse.jpg \
--checkpoint-dir checkpoint-directory \

For any problems or questions check also the notebook here https://www.tensorflow.org/hub/tutorials/tf2_arbitrary_image_stylization

Results can be like

Notes on issues:

1. If you run into an issue of tensorflow package missing compat run:
pip install tensorflow-estimator==2.1.*

2. You may need another module ‘imageio’. Then run simply
pip install imageio

You can read more about this module here https://pypi.org/project/imageio/

3. You will also need module ‘moviepy’:
pip install moviepy