Exploring Mental Prototypes by an Efficient Interdisciplinary Approach: Interactive Microbial Genetic Algorithm (IMGA)

This is the official implementation of our [Paper] Exploring Mental Prototypes by an Efficient Interdisciplinary Approach: Interactive Microbial Genetic Algorithm (IMGA).

For more information, please refer to our project website IMGA.

Approach

Requirements

Please install Pytorch, Torchvision, Psychopy, and dependencies.

pip install -r requirements.txt

Description

The generator is based on GANimation. The interface of the perceptual experiment is based on Psychopy.

The input image and the output images are 148 * 148.

datasets/results/: All the generated images can be found in test_ganimation_30.

datasets/test/imgs/: image folder.

datasets/test/aus_openface_new560.pkl: dictionary containing the action units of each image.

datasets/test/train_ids.csv: the file containing the names of the images to be used to train.

datasets/test/test_ids.csv: the file containing the names of the images to be used to test.

csv/: containing all the human feedback and the trial information.

subject/: 560 different faces are generated by activating 3 out of 16 action units. Example subject comes from MMI datasets.

*.txt: texts displayed during the perceptual experiments.

Please download the pretrained model from here and here. Then put them at ckpts/190327_160828/

To run:

IMGA.py

If you want to use a specific photograph (e.g, you are the actor)

Please replace "0.bmp" in datasets/results/test/imgs by your photo. Make sure that the size should be consistent. We suggest using a neutral face. If the face of your proposed photograph is not neutral, please follow the preprocessing instruction of our MDR project.

Citation

If you use this code or ideas from the paper for your research, please cite our paper:

@inproceedings{yan2023exploring,
  title={Exploring Mental Prototypes by an Efficient Interdisciplinary Approach: Interactive Microbial Genetic Algorithm},
  author={Yan, Sen and Soladie, Catherine and Seguier, Renaud},
  booktitle={2023 IEEE 17th International Conference on Automatic Face and Gesture Recognition (FG)},
  pages={1--8},
  year={2023},
  organization={IEEE}
}

Acknowledgement

This repository is built based on GANimation, and Openface. Sincere thanks to their wonderful works.

This project is supported by Randstad and ANR REFLETS.