Abstract:Abstract The human visual ability to recognize objects and scenes is widely thought to rely on representations in category-selective regions of visual cortex. These representations could support object vision by specifically representing objects, or, more simply, by representing complex visual features regardless of the particular spatial arrangement needed to constitute real world objects. That is, by representing visual textures. To discriminate between these hypotheses, we leveraged an image synthesis approach that, unlike previous methods, provides independent control over the complexity and spatial arrangement of visual features. We found that human observers could easily detect a natural object among synthetic images with similar complex features that were spatially scrambled. However, observer models built from BOLD responses from category-selective regions, as well as a model of macaque inferotemporal cortex and Imagenet-trained deep convolutional neural networks, were all unable to identify the real object. This inability was not due to a lack of signal-to-noise, as all of these observer models could predict human performance in image categorization tasks. How then might these texture-like representations in category-selective regions support object perception? An image-specific readout from category-selective cortex yielded a representation that was more selective for natural feature arrangement, showing that the information necessary for object discrimination is available. Thus, our results suggest that the role of human category-selective visual cortex is not to explicitly encode objects but rather to provide a basis set of texture-like features that can be infinitely reconfigured to flexibly learn and identify new object categories. Significance Statement Virtually indistinguishable metamers of visual textures, such as wood grain, can be synthesized by matching complex features regardless of their spatial arrangement (1–3). However, humans are not fooled by such synthetic images of scrambled objects. Thus, category-selective regions of human visual cortex might be expected to exhibit representational geometry preferentially sensitive to natural objects. Contrarily, we demonstrate that observer models based on category-selective regions, models of macaque inferotemporal cortex and Imagenet-trained deep convolutional neural networks do not preferentially represent natural images, even while they are able to discriminate image categories. This suggests the need to reconceptualize the role of category-selective cortex as representing a basis set of complex texture-like features, useful for a myriad of visual behaviors.

Biomimetic Construction of Category Mental Imagery Based on Recognition Mechanism of Visual Cortex of Human Brain

Biologically Inspired Model for Visual Cognition Achieving Unsupervised Episodic and Semantic Feature Learning.

A Model for Image Categorisation Based on a Biological Visual Mechanism

Brain Inspired Cognitive System for Learning and Memory

Biologically Inspired Visual Model With Preliminary Cognition and Active Attention Adjustment

Enhanced Biologically Inspired Model for Object Recognition

Reading the mind's eye: Decoding category information during mental imagery

An Image Retrieval Algorithm Based on Biomimetic Imaginal Thinking

Deep image reconstruction from human brain activity

Local Image Descriptor Inspired by Visual Cortex

A Map Construction Method Based on the Cognitive Mechanism of Rat Brain Hippocampus

Introducing Memory and Association Mechanism into a Biologically Inspired Visual Model

Texture-like representation of objects in human visual cortex

Masked Image Modeling as a Framework for Self-Supervised Learning across Eye Movements

Capturing human categorization of natural images at scale by combining deep networks and cognitive models

Multidimensional and Multiresolution Biomimetic Recognition Method and Its Application

Category Decoding of Visual Stimuli From Human Brain Activity Using a Bidirectional Recurrent Neural Network to Simulate Bidirectional Information Flows in Human Visual Cortices

A biologically inspired computational model of human ventral temporal cortex

Imagery May Arise from Associations Formed through Sensory Experience: A Network of Spiking Neurons Controlling a Robot Learns Visual Sequences in Order to Perform a Mental Rotation Task

An Improved Bioinspired Cognitive Map-Building System Based on Episodic Memory Recognition

Emotion recognition based on brain-like multimodal hierarchical perception