ckadirt's picture
Add files using upload-large-folder tool
24a6353 verified
|
Raw
History Blame Contribute Delete
1.2 kB

Scripts for extracting deep neural networks (DNNs) features from images (or text annotation), located at ./data/[dataset]/source folders. You should put the stimuli files appropriately beforehand or run the download script.

Usage (using rye environment)

If you want to extract VGG19 (for iCNN) in the Deeprecon dataset, running this command: rye run python ./analysis/0_preprocessing/iCNN_image_vgg19_feature_extraction.py ./analysis/0_preprocessing/config/Deeprecon/ImageNetTest.yaml

If you want to extract CLIP text (for Brain-Diffuser) in the Deeprecon dataset, running this command: rye run python ./analysis/0_preprocessing/BD_extract_NSD_CLIP_tex_feature.py ./analysis/0_preprocessing/config/NSD/NSD.yaml

If you want to extract CLIP vision (for Brain-Diffuser) in the NSD dataset, running this command: rye run python ./analysis/0_preprocessing/BD_extract_CLIP_vision_feature.py ./analysis/0_preprocessing/config/NSD/NSD.yaml

Note: If you want to extract CLIP features used in the Brain-Diffuser models, you need additonaly prepareversatile_diffusion at the top of this directory.