Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session A: Tuesday, August 12, 1:30 – 4:30 pm, de Brug & E‑Hall

Comparing Object Selectivity in Visual Cortex and Topographic Deep Artificial Neural Networks

Davide Cortinovis¹, Martin N Hebart², Stefania Bracci¹; ¹University of Trento, ²Justus Liebig Universität Gießen

Presenter: Davide Cortinovis

The occipitotemporal cortex (OTC) exhibits category selectivity, with specialized regions responding to specific object categories. Topographic Deep Artificial Neural Networks (TDANNs) have been proposed as mechanistic models of this spatial and functional organization. However, a direct comparison of the visual and semantic features driving functional selectivity in the two systems is lacking. We analyzed fMRI data from three participants viewing 200 images of distinct body parts and inanimate objects, and compared OTC selectivity with TDANN activations. Body-, hand-, and tool-selective regions all showed strong category preferences. TDANNs displayed similar, though weaker, selectivity with blurrier category boundaries, especially for tools. Texture scrambling revealed that TDANN selectivity partly relies on local features: body and hand selectivity persisted despite global shape disruption, while tool selectivity disappeared, possibly due to their higher similarity with the other inanimate categories. These results represent a first step toward better characterizing and comparing functional selectivity in visual cortex and topographic models.

Topic Area: Object Recognition & Visual Attention

Extended Abstract: Full Text PDF