Karim Rajaei

The paper investigates how local recurrence contributes to occluded-object recognition by analyzing neural dynamics and computational models.

The paper shows that structured scene descriptions, not just object co-occurrence, predict high-level visual cortex responses better, and language-aligned models capture this structure.

The paper shows that coherent scenes accelerate object recognition by shaping neural representations through anticipatory context, as demonstrated by EEG and multivariate pattern analysis.

This study investigates how scene context influences object recognition in humans and models, showing that vision-language models with language supervision better replicate human performance.

The paper investigates object recognition under occlusion using MEG, behavioral data, and computational models, showing that recurrent processes are essential and that a local recurrent model explains human performance.