Humans perceive the world in rich social detail. In just a fraction of a second, we not only detect objects and people in our environment, but also quickly grasp what people are doing, with whom, and why. In this talk, Isik will present research using neuroimaging (MEG, fMRI, and ECoG) and convolutional neural network models (CNNs) to characterize the computational processing stages of social vision. Isik will first describe earlier work developing and applying these methods to the case of invariant object recognition, then extend these methods to social perception, namely recognizing the actions (e.g., running or jumping) and social interactions (e.g. helping versus hindering) of others. These findings show that humans quickly form representations of actions that are invariant to viewpoint, and that CNNs provide a good model of these neural representations. Finally, Isik will discuss new results that uncover a neural correlate social interaction perception in the posterior superior temporal sulcus (pSTS), as well as ongoing work bridging high spatial and temporal resolution neuroimaging with CNNs to understand the neural computations underlying this ability.