They are salisbury novichok addict uncovering a murder suspect or even identifying a sexual predator. The research offers fresh insights into why superrecognizers excel at facial recognition.
Previous studies indicate that individuals with exceptional facial recognition skills observe more regions of the entire face compared to average individuals.
Recently, researchers have employed advanced AI techniques to reveal how this perspective enhances their capabilities.
“It’s not solely about seeing everything, it’s about using your vision intelligently,” stated the lead author of the study, Dr. James Dunn from UNSW Sydney.
In a recent article published in Proceedings of the Royal Society B: Biological Sciences, Dunn et al. highlight how they extracted eye-tracking data from a previous study involving 37 superrecognizers and 68 typical recognizers.
In their experiment, participants viewed both images of entire faces and segmented images focusing on the regions they were examining.
In this new research, the team utilized this data to reconstruct the visual information that was available to the participants’ eyes.
This “retinal information” was processed through a deep neural network (DNN), an AI system trained for facial recognition. Participants provided the AI with either a complete image of the same face they had seen or a different one.
In all instances, the AI generated a score indicating how closely the retinal information matched a given complete facial image.
The research team compared outcomes between typical participants and super-recognizers, along with data drawn from randomly chosen areas of the initial facial images.
The findings indicated that the AI system’s effectiveness improved as the visibility of the observed facial feature increased.
Moreover, across all levels of visibility, the AI performed optimally when relying on retinal data from superrecognizers.
“This suggests that variations in facial recognition capability are partly due to our active exploration and sampling of visual data, rather than just post-processing by the brain,” Dunn remarked.
The team then examined whether their findings simply indicated that superrecognizers looked at more areas of the face and gathered more data.
However, they discovered that even when the same amount of retinal information was captured, the AI performed better with data from super-recognizers.
“Their advantage lies not only in the quantity but also in the quality of information,” says Dunn. “They focus on areas that provide more identity cues, making each ‘pixel’ they select significantly more valuable for facial recognition.”
Dr. Rachel Bennett, a facial processing expert from Brunel University in London who was not involved in the study, praised the research.
“The key contribution to understanding super-recognition is that effective facial recognition isn’t only about examining specific areas or spending more time looking at the face. Super-recognizers explore not just larger areas, but also gather more advantageous data,” she asserted.
Dr. Alejandro Estudillo from Bournemouth University noted that the study was conducted by showing participants still images in highly controlled environments.
“It will be crucial to see if the same patterns emerge in more natural, dynamic contexts,” he said.
This study implies there are strategies to enhance facial recognition; however, it seems unlikely that anyone can train to become a super-recognizer.
“At present, we cannot determine if these eye movement patterns can be effectively trained,” Bennett remarked.
Dunn stated that research indicates super-recognition is likely influenced by genetics and is often inherited.
“Superrecognizers appear to instinctively identify the most crucial features. This is challenging to teach, as it differs from one face to another,” he explained.
Researchers have created a free test to help identify supercognitive traits: New South Wales face test.
Source: www.theguardian.com
