This repository contains the data and demo codes for replicating results in our paper: Horikawa and Kamitani (2017) Generic decoding of seen and imagined objects using hierarchical visual features. Nature Communications 8:15037. The generic object decoding approach enabled decoding of arbitrary object categories including those not used in model training.
The preprocessed fMRI data for five subjects (training, test_perception, and test_imagery) and visual features (CNN1-8, HMAX1-3, GIST, and SIFT) are available at figshare. The fMRI data were saved as the BrainDecoderToolbox2/bdpy format.
The unpreprocessed fMRI data is available at OpenNeuro.
For copyright reasons, we do not make the visual images used in our experiments publicly available. You can request us to share the stimulus images at https://forms.gle/ujvA34948Xg49jdn9.
Stimulus images used for higher visual area locazlier experiments in this study are available via https://forms.gle/c6HGatLrt7JtTGQk7.
Demo programs for Matlab and Python are available in code/matlab and code/python, respectively. See README.md in each directory for the details.