Masataka Sawayama

The paper prunes neural networks to isolate computational subspaces for style-invariant object recognition and tests their perceptual relevance through image synthesis.

The paper identifies recurring brain states across tasks and links their dynamics to trial-to-trial performance and individual response strategies.

The paper investigates semantic representations in body-selective brain regions by linking natural scene captions to neural responses, finding that implied body motion is the primary contributor.

This study uses the median cut algorithm to decompose color images, showing it preserves geometric structure while reducing material information, which impacts material classification but not object recognition.

This study explores the relationship between visual perception and language in material recognition using psychophysics and unsupervised learning, finding moderate correlation but structural differences in representations.

The paper investigates neural responses to color saliency differences in typical and anomalous trichromats using ERP analysis during an oddball task.

This study uses EEG to decode time-resolved neural representations during orientation ensemble perception, finding that cumulative averaging of orientation signals correlates with perceived average orientation.

This paper uses an unsupervised image generation model to discover latent features related to material perception, finding that translucency is encoded in mid-scale features.

This study constructs a database of material images with annotated visual discrimination performance and shows that material discrimination depends on illumination and geometry, with individual differences in glossiness perception.

This study investigates how color affects perception of translucent materials using real-world object photographs, finding that color improves agreement and accuracy in material classification and attribute ratings.

This study investigates how touch and vision process texture by comparing sensitivity to image statistics, finding that touch responds to lower-order statistics while vision detects higher-order ones.

This study investigates how humans perceive dynamic deformation using psychophysical experiments and a computational model of direction-selective visual neurons.