Research at the intersection of computer vision and neuroscience has revealed hierarchical correspondence between layers of deep convolutional neural networks (Dcnns) and cascade of regionsalong human ventral visual cortex. Recently, studies have uncovered emergence of human interpretable concepts within DCNNs layers trained to identify visual objects and scenes. Here, we asked whether an artificial neural network (with convolutional structure) trained for visual categorization would demonstrate spatial correspondences with human brain regions showing central/peripheral biases.
Using representational similarity analysis, we compared activations of convolutional layers of a DCNN trained for object and scene categorization with neural representations in human brain visual regions.Results reveal a brain-like topographical organization in the layers of the DCNN, such that activations of layer-units with central-bias were associated with brain regions with foveal tendencies (e.g. fusiform gyrus), and activations of layer-units with selectivity for image backgrounds were associated with cortical regions showing peripheral preference (e.g. parahippocampal cortex). the emergence of a categorical topographical correspondence between Dcnns and brain regions suggests these models are a good approximation of the perceptual representation generated by biological neural networks.Cortical regions along the ventral visual stream of the human brain (extending from occipital to temporal lobe) have been shown to preferentially activate to specific image categories 1 . For instance, while the fusiform gyrus shows specialization for faces 2 , the parahippocampal cortex (PHC) is more selective to spatial layout, places 3,4 and large-size objects 5,6 . In characterizing the functional properties of these regions, Levy and colleagues (2001) discovered distinct topographical response patterns, such that face selective regions of the fusiform gyrus showed a strong preference for central visual field, while the building selective regions of PHC exhibited a peripheral selectivity bias to images of scene and large spaces 7 . Thus, while these regions show categorical selectivity to scenes or faces, their response patterns are strongest when their preferred category is presented in a topographically favorable location in the visual field. More specifically, the face selective voxels in the fusiform gyrus have a stronger response when faces are presented centrally; whereas scene-selective voxels show stronger activity to space features in the periphery 7-13 .These topographical preferences raise questions regarding the origin of this functional organizing principles: does the way we look at faces and scenes in our natural visual world account for this bias? We most often fixate on faces bringing face-related information into our central, high acuity fovea to extract subtle visual features like facial expressions [14][15][16] . Places, on the other hand, are used for navigation, extending all around the visual field, thus we more readily percei...