This dataset is used to train and evaluate the kwyk network. We are publishing it in the form of a datalad dataset.
The main folder includes two subfolders rawdata
and tfrecords
. The rawdata
folder includes input images with the freesurfer segementation files. The tfrecords
has the same files as in rawdata
folder but they are saved as pair of (image, segmetation)
in the form of tfrecords.
The data in tfrecords
folder is separated to train and test.
To download the data, you need to do following steps:
We highly recommend to create a separate environment for accessing the data. You need to install datalad, git-annex and rclone. Plesae check here to insttall datalad and it's dependencies based on your operating system. For linux systems you can install datalad and dependencies using conda like below.
conda create -n datalad python=3
conda activatet datalad
conda install -c conda-forge rclone
conda install -c conda-forge git-annex
conda install -c conda-forge datalad
Note: If you are using the datalad for the first time, remember to configure it.
install git-annex-remote-rclone:
find path to the bin directory of your conda env with which python
.
which python | grep -o ".*bin"
copy the git-annex-remote-rclone to the bin directory.
wget https://raw.githubusercontent.com/DanielDent/git-annex-remote-rclone/master/git-annex-remote-rclone \
-P </path/to/conda/env/bin>
chmod +x </path/to/conda/env/bin>/git-annex-remote-rclone
Now you need to setup the rclone remote.
rclone config
use n
to setup a new remote.
Except for below fields use default for everything else:
Name:gd-sig
Type of storage: 13 (Google Drive)
Scope: 1 (full access)
Auto config: No
You’ll see something like this after auto config:
Copy the link given in the terminal. On your local computer, open any browser and go to the link. Log into Google Drive with the email you registered MIT G suite with and allow access to your Google account.
Copy the verification code that pops up.
Go back to your terminal. Paste the code there and select ‘yes’ for team drive.
Enter ID of the drive you want to add, enter yes and you’re done!
Clone this repository using datalad. Note that by cloning the repository you are getting the metadata of the files rather than the actual content.
datalad clone https://github.com/sensein/kwyk-data
Your sample output of datalad clone should look like this:
[INFO ] Remote origin not usable by git-annex; setting annex-ignore
[INFO ] access to 1 dataset sibling gdsig not auto-enabled, enable with:
| datalad siblings -d "/home/testdatalad" enable -s gd-sig
install(ok): /home/testdatalad (dataset)
Use the 3rd line as the command after cding into cloned repo
cd kwyk-data
datalad siblings -d "/home/nlo/testdatalad" enable -s gdsig # copy line from output
You can use datalad get <path/to/data file/>
to download the data.
To download the whole rawdata folder:
datalad get rawdata
To download the whole tfrecords:
datalaad get tfrecords
To download a single file, simply pass the path of the file to datalad get
.
datalad get rawdata/rawdata/pac_0_orig.nii.gz