Project page: https://sites.google.com/view/dnn2brainfunction/home
Code: https://github.com/cvai-repo/dnn2brain-function
Unzip the data files provided here.
Data is organized into subfolders contatining representational dissimilarity matrices of DNNs/ROIs/searchlights. Please find detailed description of each file below
- data/ROIfiles_Labeling.txt : Text file containing list of ROIs
- data/mask.nii : Map containing brain mask in MNI space
- data/kastner_ROIs_RDMs_pearson:
- Folder containing ROI RDMs
- RDMs in are in ROI_name.mat (e.g. V1v.mat) format where ROI_name is the name of ROI mentioned in ROI label file.
- ROI_name.mat : has a variable name 'subject_rdm' which is a 16x50x50 (num_subjects x num_conditions x num_conditions) matrix
- data/RDM_taskonomy_bonner50 :
- Folder containing Taskonomy DNN RDMs
- RDMs are in task_encoder_layer.mat (e.g. autoencoder_encoder_block4.mat) file where task is the name of taskonomy tasks, and layer represents the layer output from which the RDMs where created.
- task_encoder_layer.mat: has a variable name 'rdm' which is a 50x50 (num_conditions x num_conditions) matrix
- data/sl_rad=1__max_blk_edge=2:
- Folder containing searchlight RDMs
- all_sub_slrdms.npy: Searchlight RDMs for all the subject is a matrix of dimensions 16 x 30679 x 50 x 50 (num_subjects x num_searchlights x num_conditions x num_conditions)
- lnc.nii : lower noise ceiling map in MNI space
- unc.nii : upper noise ceiling map in MNI space
- noise_ceiling.pkl: file containing lower and upper noise ceiling in 1-D np.array with dimension = 30679 (num_searchlights)
The data provided here is originally from Bonner and Epstein, PNAS 2017. Please cite both the papers below if you use the data.
1. Bonner, M. F., & Epstein, R. A. (2017). Coding of navigational affordances in the human visual system. Proceedings of the National Academy of Sciences, 114(18), 4793-4798.
2. Dwivedi, K., Bonner, M. F., Cichy, R. M. & Roig, G. Unveiling functions of the visual cortex using task-specific deep neural networks. bioRxiv 2020.11.27.401380 (2020) doi:10.1101/2020.11.27.401380.
