Frank Tong

This study shows that training DNNs with natural occluders improves their robustness to real-world occlusion, unlike artificial occluders which do not replicate human-like performance.

This study shows that training convolutional neural networks with blurry images improves their performance in predicting neural responses and aligns them better with human vision.

This study compares CNN models to understand non-linear processing in macaque V1, finding that few non-linear stages can explain V1 responses.

This study trains CNNs with blurry to clear images for face and object recognition, finding that face networks retain blur robustness while object networks do not, suggesting differences in processing strategies.

This study shows that noise-trained deep neural networks better match human performance and neural responses to noisy images than standard DNNs.

The paper uses population receptive field modeling and high-field fMRI to resolve spatial profiles of figure-ground modulation mechanisms in human V1, revealing feedback from higher visual areas.

This study investigates how transcranial magnetic stimulation (TMS) over early visual cortex affects short-term memory precision and guess rate, finding enhanced memory precision at targeted retinotopic locations and increased guessing with early TMS pulses.