Counting 3,834 Big Data & Machine Learning Frameworks, Toolsets, and Examples...
Suggestion? Feedback? Tweet @stkim1

FeatureTransferApp

Image editing based on a Variational Autoencoder.

For usage and explanation consult this youtube video:

Video

Installation

Install guide for ubuntu based systems. Windows/Mac can deviate slighty.

Create and activate virtual enviroment (optional but recommended)

This will keep changes to your python installation contained in the virtual enviroment.

python3 -m venv venv-feature-transfer
source venv-feature-transfer/bin/activate

Install

Some libraries used require additional build tools. Install them with

sudo apt-get install build-essential cmake pkg-config
sudo apt-get install libx11-dev libatlas-base-dev
sudo apt-get install libgtk-3-dev libboost-python-dev

Install python packages listed with

pip3 install -r requirements.txt

Execute

Start the program with

python3 main.py

If you used a virtual environment to install the dependencies, you need to activate it with source venv-feature-transfer/bin/activate before each program start.

Add your own Images

To add your own images, place them in the directory additional_images. All common image file formats are accepted. Images should only contain one face. On program startup, the images are preprocessed. Sucessfully processed images are placed in the directory additional_images_preprocessed. If no face / more than one face is found in the image, it is ignored.