“I Am Told to Be Happy”: An Exploration of Deep Learning in Affective Colormaps in Industrial Tomography

Humans show different emotions in response to variant colormaps when facing visual presentations. The affect—colormap relationship thus becomes an important factor in human-in-the-loop systems. In this paper, we explore how to effectively exploit deep learning in affective colormaps within the domain of industrial tomography. Eleven pervasively used colormaps were picked as the stimuli, followed by a user study which gathered data on the human affect of each colormap as well as benchmarking our initial dataset. The affect was encoded into an emotional model over two dimensions; valence (positive—negative), and arousal (exciting—calm). Our proposed convolutional neural network (CNN) consisting of 10 layers reached high recognition and prediction accuracy in the colormap—affect relationship. The obtained results affirmed our exploration, which could in future assist developers to construct more intelligent and reliable human-computer interaction (HCI) systems.

Paper

Full text

PDF

“I Am Told to Be Happy”: An Exploration of Deep Learning in Affective Colormaps in Industrial Tomography

Semantic Scholar · Engineering · 2021

Abstract

Humans show different emotions in response to variant colormaps when facing visual presentations. The affect—colormap relationship thus becomes an important factor in human-in-the-loop systems. In this paper, we explore how to effectively exploit deep learning in affective colormaps within the domain of industrial tomography. Eleven pervasively used colormaps were picked as the stimuli, followed by a user study which gathered data on the human affect of each colormap as well as benchmarking our initial dataset. The affect was encoded into an emotional model over two dimensions; valence (positive—negative), and arousal (exciting—calm). Our proposed convolutional neural network (CNN) consisting of 10 layers reached high recognition and prediction accuracy in the colormap—affect relationship. The obtained results affirmed our exploration, which could in future assist developers to construct more intelligent and reliable human-computer interaction (HCI) systems.

Similar papers

© 2026 NYSGPT2525 LLC