Koustav Banerjee Presents on the Brain's Perception of Implied Motion

Koustav Banerjee, graduate student, recently presented a poster, titled “Decoding Implied Motion Magnitude Using fMRI Activity,” at the 2026 Cognitive Computational Neuroscience (CCN) conference. The CCN conference unites researchers in cognitive science, neuroscience, and artificial intelligence (AI) to discover computational underpinnings of cognition.

Banerjee’s research used the large-scale fMRI Natural Scenes Dataset (NSD) to study how the brain represents motion from photos, like an image of a surfer riding a wave, even though the photo is static. His findings revealed that implied motion is processed across a broader network of the brain than past works suggest, with neural activity encoding the magnitude of perceived movement precisely. He combined the latest mechanistic interpretable AI methods like foundation-models and sparse autoencoders to reveal specific image features that are encoded by the brain to drive this prediction of motion, offering a keen insight about what information the visual system uses for this automatic process.