Deep learning of shared perceptual representations for familiar and unfamiliar faces: Reply to commentaries

Humans' extreme face recognition abilities challenge the well-established familiarity effect

Classification performance is better for learned than unlearned stimuli. This was also reported for faces, where identity matching of unfamiliar faces is worse than for familiar faces. This familiarity advantage led to the conclusion that variability across appearances of the same identity is partly idiosyncratic and cannot be generalized from familiar to unfamiliar identities. Recent advances in machine vision challenge this claim by showing that the performance for untrained (unfamiliar) identities reached the level of trained identities as the number of identities that the algorithm is trained with increases. We therefore asked whether humans who reportedly can identify a vast number of identities, such as super recognizers, may close the gap between familiar and unfamiliar face classification. Consistent with this prediction, super recognizers classified unfamiliar faces just as well as typical participants who are familiar with the same faces, on a task that generates a sizable familiarity effect in controls. Additionally, prosopagnosics' performance for familiar faces was as bad as that of typical participants who were unfamiliar with the same faces, indicating that they struggle to learn even identity-specific information. Overall, these findings demonstrate that by studying the extreme ends of a system's ability we can gain novel insights into its actual capabilities.

A narrow band of image dimensions is critical for face recognition

A key challenge in human and computer face recognition is to differentiate information that is diagnostic for identity from other sources of image variation. Here, we used a combined computational and behavioural approach to reveal critical image dimensions for face recognition. Behavioural data were collected using a sorting and matching task with unfamiliar faces and a recognition task with familiar faces. Principal components analysis was used to reveal the dimensions across which the shape and texture of faces in these tasks varied. We then asked which image dimensions were able to predict behavioural performance across these tasks. We found that the ability to predict behavioural responses in the unfamiliar face tasks increased when the early PCA dimensions (i.e. those accounting for most variance) of shape and texture were removed from the analysis. Image similarity also predicted the output of a computer model of face recognition, but again only when the early image dimensions were removed from the analysis. Finally, we found that recognition of familiar faces increased when the early image dimensions were removed, decreased when intermediate dimensions were removed, but then returned to baseline recognition when only later dimensions were removed. Together, these findings suggest that early image dimensions reflect ambient changes, such as changes in viewpoint or lighting, that do not contribute to face recognition. However, there is a narrow band of image dimensions for shape and texture that are critical for the recognition of identity in humans and computer models of face recognition.

Stable individual differences in unfamiliar face identification: Evidence from simultaneous and sequential matching tasks

Matching identity in images of unfamiliar faces is difficult: Images of the same person can look different and images of different people can look similar. Recent studies have capitalized on individual differences in the ability to distinguish match (same ID) vs. mismatch (different IDs) face pairs to inform models of face recognition. We addressed two significant gaps in the literature by examining the stability of individual differences in both sensitivity to identity and response bias. In Study 1, 210 participants completed a battery of four tasks in each of two sessions separated by one week. Tasks varied in protocol (same/different, lineup, sorting) and stimulus characteristics (low vs. high within-person variability in appearance). In Study 2, 148 participants completed a battery of three tasks in a single session. Stimuli were presented simultaneously on some trials and sequentially on others, introducing short-term memory demands. Principal components analysis revealed two components that were stable across time and tasks: sensitivity to identity and bias. Analyses of response times suggest that individual differences in bias reflect decision-making processes. We discuss the implications of our findings in applied settings and for models of face recognition.

What happens to our representation of identity as familiar faces age? Evidence from priming and identity aftereffects

Matching identity in images of unfamiliar faces is error prone, but we can easily recognize highly variable images of familiar faces - even images taken decades apart. Recent theoretical development based on computational modelling can account for how we recognize extremely variable instances of the same identity. We provide complementary behavioural data by examining older adults' representation of older celebrities who were also famous when young. In Experiment 1, participants completed a long-lag repetition priming task in which primes and test stimuli were the same age or different ages. In Experiment 2, participants completed an identity after effects task in which the adapting stimulus was an older or young photograph of one celebrity and the test stimulus was a morph between the adapting identity and a different celebrity; the adapting stimulus was the same age as the test stimulus on some trials (e.g., both old) or a different age (e.g., adapter young, test stimulus old). The magnitude of priming and identity after effects were not influenced by whether the prime and adapting stimulus were the same age or different age as the test face. Collectively, our findings suggest that humans have one common mental representation for a familiar face (e.g., Paul McCartney) that incorporates visual changes across decades, rather than multiple age-specific representations. These findings make novel predictions for state-of-the-art algorithms (e.g., Deep Convolutional Neural Networks).

Face Recognition by Humans and Machines: Three Fundamental Advances from Deep Learning

Deep learning models currently achieve human levels of performance on real-world face recognition tasks. We review scientific progress in understanding human face processing using computational approaches based on deep learning. This review is organized around three fundamental advances. First, deep networks trained for face identification generate a representation that retains structured information about the face (e.g., identity, demographics, appearance, social traits, expression) and the input image (e.g., viewpoint, illumination). This forces us to rethink the universe of possible solutions to the problem of inverse optics in vision. Second, deep learning models indicate that high-level visual representations of faces cannot be understood in terms of interpretable features. This has implications for understanding neural tuning and population coding in the high-level visual cortex. Third, learning in deep networks is a multistep process that forces theoretical consideration of diverse categories of learning that can overlap, accumulate over time, and interact. Diverse learning types are needed to model the development of human face processing skills, cross-race effects, and familiarity with individual faces.