petrus-mikkola / GANPPBO

Research project combining GANSpace and Projective Preferential Bayesian Optimization technique for modelling user's preferences in generated images

GANspace + PPBO project

Idea

Probabilistic Interactive User Model for Interactive AI

Based on 2 projects:

• GANSpace
• Projective Preferential Bayesian Optimization (PPBO)

Idea Formulation: Adjust image generation model (GAN) to follow user’s preferences using Bayesian Optimization.

Set Up

1. Install anaconda or miniconda
2. Create environment: conda create -n ganppbo python=3.7
3. Activate environment: conda activate ganppbo
4. Open project directory: cd GANPPBO
5. Install dependencies: conda env update -f environment.yml --prune
6. Setup submodules: git submodule update --init --recursive
7. Activate widget extension: jupyter nbextension enable --py widgetsnbextension

Original project info

Currently, the code of ganspace projects is committed to this repository. Later ganspace is going to be transformed into submodule.

Original repositories:

• https://github.com/harskish/ganspace
• https://github.com/AaltoPML/PPBO

Origial README files:

• GANSpace README file
• PPBO README file