Visual attention : a study on image sentiment and emotional priority

Abstract

Emotions on the image influences visual perception of the viewers. Stimuli such as laughing faces and crying child attracts human attention than neutral images with no emotional stimuli. This research is to evaluate the relationship between the sentiment of the image and visual attention of the user based on the emotional properties of the image. In an attempt to understand the effects of image attributes on predict, three different convolutional neural network models were designed was implemented in Python with Keras and Tensorflow. This, together with implementations of the saliency map to identify the emotional prioritization on the images. As the results of using different networks was compared and predicted that image set with positive emotions has more attention than images with negative and neutral emotions. The results were also compared to previous attempts to classify the same data to evaluate the method as a whole.