Cortically-Coupled Generative Adversarial Network for Target Image Retrieval in Rapid Image Search

Rapid growth in the multimedia and healthcare domain resulted in a tremendous increase in visual data. It has become difficult to access this visual data due to its huge volume and unstructured nature. Over the past few decades, computer vision research is focused on finding a smart way to retrieve the visual data of interest in rapid serial visual presentation (RSVP) events and understanding the brain sensory stimuli response to such events. In this paper, the focus is on developing the system that can relate a brain state to target identification and analysis in an RSVP. In this research, the P300 event occurred due to the shift in attention is analyzed and captured using the electroencephalogram (EEG). A model called Cortically-Coupled Generative Adversarial Network is proposed using this analysis. This model identifies and retrieves the target image in RSVP events. The evaluation of the proposed model demonstrates the combination of EEG signals and cortically-coupled GAN could effectively use to develop a smart way to retrieve the visual data of interest.