Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evaluation and Comparison of Emotionally Evocative Image Augmentation Methods

Jan Ignatowicz, Krzysztof Kutt, Grzegorz J. Nalepa

TL;DR

An approach to creating stimulus datasets for affective computing using generative adversarial networks (GANs) is explored, highlighting promising advances in the generation of emotionally evocative synthetic images and suggesting significant potential for future research and improvements.

Abstract

Experiments in affective computing are based on stimulus datasets that, in the process of standardization, receive metadata describing which emotions each stimulus evokes. In this paper, we explore an approach to creating stimulus datasets for affective computing using generative adversarial networks (GANs). Traditional dataset preparation methods are costly and time consuming, prompting our investigation of alternatives. We conducted experiments with various GAN architectures, including Deep Convolutional GAN, Conditional GAN, Auxiliary Classifier GAN, Progressive Augmentation GAN, and Wasserstein GAN, alongside data augmentation and transfer learning techniques. Our findings highlight promising advances in the generation of emotionally evocative synthetic images, suggesting significant potential for future research and improvements in this domain.

Evaluation and Comparison of Emotionally Evocative Image Augmentation Methods

TL;DR

An approach to creating stimulus datasets for affective computing using generative adversarial networks (GANs) is explored, highlighting promising advances in the generation of emotionally evocative synthetic images and suggesting significant potential for future research and improvements.

Abstract

Experiments in affective computing are based on stimulus datasets that, in the process of standardization, receive metadata describing which emotions each stimulus evokes. In this paper, we explore an approach to creating stimulus datasets for affective computing using generative adversarial networks (GANs). Traditional dataset preparation methods are costly and time consuming, prompting our investigation of alternatives. We conducted experiments with various GAN architectures, including Deep Convolutional GAN, Conditional GAN, Auxiliary Classifier GAN, Progressive Augmentation GAN, and Wasserstein GAN, alongside data augmentation and transfer learning techniques. Our findings highlight promising advances in the generation of emotionally evocative synthetic images, suggesting significant potential for future research and improvements in this domain.

Paper Structure

This paper contains 17 sections, 6 figures, 4 tables.

Table of Contents

  1. Introduction and Motivation
  2. Generative Models
  3. Datasets
  4. Generation of Affective Pictures
  5. Preprocessing
  6. Generative models
  7. Evaluation
  8. Experiments summary
  9. Generation Results
  10. Classification Results
  11. DALL-E
  12. Summary and Future Work
  14. An example appendix
  15. Example of a sub-heading within an appendix
  16. ...and 2 more sections

Figures (6)

  • Figure 1: [a] Russel's circumplex model of emotions val-aro-diagram, [b] Pseudo-randomly selected OASIS images (based on oasis)
  • Figure 2: [a] Normalized ratings for images from all datasets, [b] Valence-arousal space divided into 13 categories (cf. Fig. \ref{['fig:all_data_valence_arousal_and_classified']}[a])
  • Figure 3: [a] CIFAR-10 fakes by DCGAN D., [b] A. AFFECTIVE fakes by DCGAN D.
  • Figure 4: [a] CIFAR-10 fakes by DCGAN SN., [b] A. AFFECTIVE fakes by DCGAN SN., [c] CIFAR-10 fakes by CGAN D., [d] AFFECTIVE fakes by CGAN D., [e] AFFECTIVE fakes by PAGAN D., [f] AFFECTIVE fakes by WGAN GP. D.
  • Figure 5: Models best FID scores on A. AFFECTIVE dataset
  • ...and 1 more figures