This Python package provides interfaces to datasets published from Kamitani Lab, Kyoto Univ and ATR.
$ pip install git+https://github.com/KamitaniLab/brain-decoding-datasets.git
$ python -m bdds.downloadall <dataset name> [--output <output directory>]
Example:
$ python -m bdds.downloadall handshape --output data/handshape
$ python -m bdds.downloadall god --output data/god
$ python -m bdds.downloadall decodeddnn --output data/decodeddnn
import bdds
# Hand shape decoding dataset
dataset_handshape = bdds.HandShapeDecoding('data/handshape')
data_handshape_s1 = dataset_handshape.get(mode='fmri', subject='S1') # Return fMRI data as a bdpy dataset
# GOD fMRI dataset
dataset_god = bdds.GenericObjectDecoding('data/god')
data_god_s1 = dataset_god.get(mode='fmri', subject='Subject1') # Return fMRI data as a bdpy dataset
# Decoded DNN features
dataset_dnn = bdds.DecodedDNN('data/decodeddnn')
decoded_feature = dataset_dnn.get(mode='decoded', subject='S1', net='AlexNet', layer='fc8') # Return features as (a list of) numpy arrays.
If data files are missing in your local filesystem, the dataset instance asks you whether to download the files from online repositories or not.
Example:
>>> import bdds
>>> dataset_handshape = bdds.HandShapeDecoding('data/handshape')
>>> data_handshape_s1 = dataset_handshape.get(mode='fmri', subject='S1')
Data file is missing. Download? (y/[n])y
Downloading https://ndownloader.figshare.com/files/12227786
Saved data/handshape/S1.h5
>>>
See demo.ipynb for more details.
When the dataset instance is initialized with with auto_download=True
option, it downloads automatically missing files from data repositories.
Example:
>>> import bdds
>>> dataset_handshape = bdds.HandShapeDecoding('data/handshape', auto_download=True)
>>> data_handshape_s1 = dataset_handshape.get(mode='fmri', subject='S1')
Downloading https://ndownloader.figshare.com/files/12227786
Saved data/handshape/S1.h5
>>>
- Generic object decoding ('god') #fMRI #DNN
- DNN features decoded from human brain activity ('decodeddnn') #fMRI #DNN
- Hand shape decoding dataset ('handshape') #fMRI