Temporal adaptation during visual processing Insights from humans and vision models

The world around us is inherently dynamic and our ability to adapt to continuous change is fundamental to human perception. One mechanism thought to support this ability is temporal adaptation, referring to the adjustment of neuronal responses to the temporal structure of incoming visual inputs. While temporal adaptation has been extensively studied, open questions remain regarding how it shapes neural responses and influences perception. In this thesis, we combine neural and behavioral measurements to examine how temporal adaptation shapes the representation of dynamic visual inputs across cortical regions and how these adaptive processes influence human ability to recognize targets in degraded environments. In addition, we use computational models to identify potential mechanisms underlying temporal adaptation, thereby providing insight into how neural circuits implement adaptive processing over time. Our results demonstrate that temporal adaptation shapes neural responses across the visual hierarchy in an area- and stimulus-dependent manner, suggesting that temporal adaptation is tuned to the functional specializations of different visual regions. At the perceptual level, we show that temporal adaptation enhances object recognition under challenging conditions such as visual noise, highlighting its role in stabilizing perception when inputs are degraded. Moreover, through computational modeling, we reveal functional differences across distinct temporal adaptation mechanisms, suggesting that adaptation arises from an interplay between bottom-up and top-down processes that engage both individual neurons and broader neural circuits. Overall, this work advances our understanding of how the visual system exploits temporal structure in the environment to construct stable and efficient representations of dynamic sensory input.