atrudel / audra_extended

Extension of the AuDrA model, originally built by researchers to rate the creativity of human drawings

AuDrA Extended

Extension of the AuDrA model, originally built by a team of researchers to rate the creativity of human drawings, as published in this paper:

Patterson, J. D., Barbot, B., Lloyd-Cox, J., & Beaty, R. E. (2023). AuDrA: An automated drawing assessment platform for evaluating creativity. Behavior Research Methods. https://doi.org/10.3758/s13428-023-02258-3

We worked on their existing code base, available on OSFHome. The orignal code is contained in the first commit of this repository and can be found on the original_AuDrA branch.

Installation

The conda environment used in this repository is made to be run locally, on CPU.

1- Make sure you have Python and conda installed (with Anaconda for example)

2- Clone the repository

git clone https://github.com/atrudel/AuDrA_extended.git

3- Make sure you are at the root of the repository

cd AuDrA_extended

2- Run the setup script that will

• Create the conda environment
• Download the training data and store it in a Drawings folder

sh setup.sh

Visualize AuDrA features

Features outputted by the penultimate layer of the pre-trained AuDra Resnet are extracted using all images in the dataset, then a PCA is performed for visualization.

A Flask app allows easy visualization of the first 3 principal components.

From the root of the repository:

python inerpretation.py

Then click on the link displayed in the terminal to open the Flask app in your browser.