Use DNNs to build encoding models of EEG visual responses.

Human Things explores whether human similarity judgments can improve machine vision embeddings by fine-tuning ResNet-50 models on THINGS object images and human odd-one-out similarity data.