Code for
Straub, D., & Rothkopf, C. A. (2021). Looking for image statistics: active vision with avatars in a naturalistic virtual environment. Frontiers in Psychology, 12, 431, Research Topic: Modeling Virtual Humans for Understanding the Mind
https://www.frontiersin.org/articles/10.3389/fpsyg.2021.641471/full
This repository consists of three main parts: simulating virtual agents, retina projection & image statistics. Before running the code, you need a Python 3 environment with all packages mentioned in requirements.txt. An easy way to set this up is using virtual environments. After cloning the repo and switching into the directory, run
chmod +x setup-env.sh
./setup-env.sh
to set up a virtual environment, update pip and install all requirements.
To download the image dataset, go to https://osf.io/5xqwc/ and download the individual zip files, saving them to data/images/human/ and extract them, such that the images from each visual field position are in their own directory, e.g. data/images/human/ecc0_polar0/.
Given the positions and orientations of a human participant, the notebook VirtualAgents creates viewing directions of the three virtual agents (straight, down and random) used in the paper. The results are saved in data/virtual-agents. The image datasets for the virtual agents were then generated from these positions and directions using a custom Unity environment. Due to its large size, it is only available upon request: straub@psychologie.tu-darmstadt.de
To project the images onto tangential planes of an idealized retina, simply run
python project.py data/images/human
The transformed images will then be in data/images/human-transformed. The script reads the information about the camera's properties and the visual field positions from info.txt in the data directory.
A minimal working example of the image analysis methods used in the paper can be run via
python main.py