Deep Neural Networks predict Hierarchical Spatio-temporal Cortical Dynamics of Human Visual Object Recognition

TL;DR

This study demonstrates that deep neural networks trained on real-world visual tasks can accurately model the spatio-temporal hierarchical processing stages of human visual object recognition as observed through MEG and fMRI data.

Contribution

The paper shows that DNNs trained on real-world categorization tasks replicate the spatio-temporal dynamics of human visual processing, highlighting the importance of training data and architecture.

Findings

DNNs capture early visual and dorsal/ventral stream processing stages.

Training on real-world data is crucial for hierarchical brain-like representations.

Model architecture influences the alignment with human visual processing.

Abstract

The complex multi-stage architecture of cortical visual pathways provides the neural basis for efficient visual object recognition in humans. However, the stage-wise computations therein remain poorly understood. Here, we compared temporal (magnetoencephalography) and spatial (functional MRI) visual brain representations with representations in an artificial deep neural network (DNN) tuned to the statistics of real-world visual recognition. We showed that the DNN captured the stages of human visual processing in both time and space from early visual areas towards the dorsal and ventral streams. Further investigation of crucial DNN parameters revealed that while model architecture was important, training on real-world categorization was necessary to enforce spatio-temporal hierarchical relationships with the brain. Together our results provide an algorithmically informed view on the spatio-temporal dynamics of visual object recognition in the human visual brain.