Supporting Online Material Materials and Methods Figs. S1 to S10 References Movies S1 to S3 Functional Compartmentalization and Viewpoint Generalization within the Macaque Face-processing System

By A R Simard, D. Soulet, G. Gowing, J. Julien, S. Rivest, B Ajami, J. Bennett, C. Krieger, W. Tetzlaff, F. M. Rossi and 10 more

692 citations recorded by the source corpus.

Tuned for Faces The temporal lobe of macaques' brains contains six patches of face-selective cortex. This observation has prompted systems neuroscientists to ask, why so many and what do they do? Freiwald and Tsao (p. 845; see the Perspective by Connor) targeted four of these regions for single-unit recordings and found that the different face-selective patches in macaques have independent functions. The areas where earliest processing occurred were most sharply tuned for individual views and least sharply tuned for identity. The mid-level area was more sharply tuned for identity, and the highest processing stage was strongly tuned for identity in a strikingly view-invariant way. These results yield fundamental insights into the computational process of object recognition, the functional organization of the brain, and how representations are transformed through processing hierarchies. Recognition of faces from different viewpoints involves several distinct stages of neural processing. Primates can recognize faces across a range of viewing conditions. Representations of individual identity should thus exist that are invariant to accidental image transformations like view direction. We targeted the recently discovered face-processing network of the macaque monkey that consists of six interconnected face-selective regions and recorded from the two middle patches (ML, middle lateral, and MF, middle fundus) and two anterior patches (AL, anterior lateral, and AM, anterior medial). We found that the anatomical position of a face patch was associated with a unique functional identity: Face patches differed qualitatively in how they represented identity across head orientations. Neurons in ML and MF were view-specific; neurons in AL were tuned to identity mirror-symetrically across views, thus achieving partial view invariance; and neurons in AM, the most anterior face patch, achieved almost full view invariance.

Subfields: Cognitive Neuroscience

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

Citations in the corpus

View-tolerant face recognition and Hebbian learning imply mirror-symmetric neural tuning to head orientation (2016, positive): “Furthermore, spatial position and size invariance increase from ML/MF to AL, and further to AM [16].”
Similarity of Neural Network Representations Revisited (2019, neutral): “…representational similarity analysis (Kriegeskorte et al., 2008a; Edelman, 1998) to compare representations across brain areas (Haxby et al., 2001; Freiwald & Tsao, 2010), individuals (Connolly et al., 2012), species (Kriegeskorte et al., 2008b), and behaviors (Elsayed et al., 2016), as well as…”
How does the brain solve visual object recognition? (2012, neutral): “Similarly, studies of fMRI-targeted clusters of IT neurons suggest that IT subpopulations can support other object recognition tasks such as face detection and face discrimination over some identity-preserving transformations (Freiwald and Tsao, 2010).”
Foundations of Vision (2018, background): “While most face-selective neurons in the monkey appear to be tuned in a view-specific manner, a subset have been found to be view-invariant, especially in more anterior regions of the temporal lobe (Freiwald & Tsao, 2010; Perrett et al., 1991).”
Intelligent Information Loss: The Coding of Facial Identity, Head Pose, and Non-Face Information in the Macaque Face Patch System (2015, methodology): “In agreement with Freiwald and Tsao (2010) we see an increase in pose-invariant identity information from ML/MF to AL to AM (Figs.”
Performance-optimized hierarchical models predict neural responses in higher visual cortex (2014, positive): “S2B), which correspond intuitively to architecturally specialized subregions like those observed in the ventral visual stream (13, 32).”
The Invariance Hypothesis Implies Domain-Specific Regions in Visual Cortex (2014, neutral): “[18] Winrich A.”
Deep neural networks: a new framework for modelling biological vision and brain information processing (2015, neutral): “Individual faces, for example, become more clearly distinguishable in monkey neuronal population codes at latencies that exceed the 100 ms or so it takes the feedforward sweep to reach IT (Sugase et al. 1999; Freiwald & Tsao 2010).”
Cognitive computational neuroscience (2018, positive): “A brain-computational model of face recognition(27) will have to explain the spatial clusters of face-selective units and the selectivities and invariances observed with fMRI(28,29) and invasive recordings.(25,30) Cognitive neuroscience has mapped the global functional layout of the human and nonhum”
Deep Learning in Spiking Neural Networks (2018, positive): “…accepted that the ability of the brain to recognize complex visual patterns or identify auditory targets in a noisy environment is a result of several processing stages andmultiple learningmechanisms embedded in deep spiking networks (Felleman & Van Essen, 1991; Freiwald & Tsao, 2010; Serre, 2014).”
Efficient and robust analysis-by-synthesis in vision : A computational framework , behavioral tests , and modeling neuronal representations (2015, methodology): “our generative model, which mimicked the FV image dataset from Freiwald and Tsao (2010). Our dataset contained 7 head poses of 25 different identities under one lighting condition (e.”
Multi-View Perceptron: a Deep Model for Learning Face Identity and View Representations (2014, background): “Recent studies [5, 19] discovered that primate brain has a face-processing network, where view and identity are processed by different neurons.”