1. Get Started¶

1.1. Requirements¶

python==3.6.x/3.7.x
1.16.5=<numpy<1.18
nibabel>=2.4.1
nipy>=0.4.2
h5py>=2.9.0
scikit-image>=0.15.0
scikit-learn>=0.19.1
scipy>=1.3.1
torch>=1.3.0
torchvision>=0.4.2
pillow>=6.0.0
opencv-python>=4.1.0.25
matplotlib>=2.2.2
cifti>=1.1
resampy
soundfile
six

A easy way to get most of these requirements is to install a python distribution (e.g. Anaconda is recommended).

After preparing these requirements, you can install DNNBrain by one of the following installation schemes. Finally, do not forget to read the Configuration and Download DNN parameters.

Tip: If the installation scheme you choose requires configuring environment variables, and you don’t know how to do it. Please skip to Configuration

1.2. Install from source¶

Get source code
Method1: Download source archive file (zip file) from GitHub and unpack it.
Method2: Clone the DNNBrain repostitory by executing: git clone https://github.com/BNUCNL/dnnbrain.git
Execute setup.py
Change to source directory (it should have the files README.md and setup.py).
Run python setup.py install

If you don’t have permission to install DNNBrain to the directory used
by the above-mentioned command, you can install it to “the Python user
install directory” by using –-user option:
python setup.py install --user
Please ensure that “bin” or “Scripts” directory (named as
path_to_dnnbrain_bin) in “the Python user install directory” is
added to the environment variable PATH, otherwise, DNNBrain’s
command-line interfaces can’t be searched by the shell.

In addition, you can also install DNNBrain to the directory specified by yourself through the –-prefix or –-home options to setup.py.

Note: If you want to install a package out of the standard python site-packages directory, you have to update two environment variables (PATH and PYTHONPATH) for your preferred directory. See Installing Python Modules for further details.

1.3. Install portable edition¶

This installation scheme allows you put DNNBrain anywhere you like as long as setting the two environment variables (PATH and PYTHONPATH) appropriately. The scheme is described as below:

Get source code through methods mentioned at step 1 of Install from source.
Add source directory (named as path_to_dnnbrain) to PYTHONPATH and “bin” directory under it (named as path_to_dnnbrain_bin) to PATH.

1.4. Configuration¶

No matter what installation scheme you choose, you have to set the environment variable DNNBRAIN_DATA whose value is the absolute path of a directory that used to store DNNBrain’s data, such as test data and pretrained parameters of DNNs.

Note: If you work with DNNBrain on a public platform, such as a Linux server, DNNBRAIN_DATA should be configured by the administrator of the platform.

Note: The updates for PATH and PYTHONPATH are optional according to the installation scheme’s requirement you choose.

Assuming the value of DNNBRAIN_DATA is path_to_dnnbrain_data.
Assuming the value which should be added to PATH is
path_to_dnnbrain_bin.
Assuming the value which should be added to PYTHONPATH is
path_to_dnnbrain.

Linux | MAC OS X¶

Open a terminal window and make sure that you know which “shell” you are running (If not sure, try running the command printenv SHELL).

For bash shell, run the following commands to set environment variables:

echo 'export DNNBRAIN_DATA=path_to_dnnbrain_data' >> ~/.bashrc
echo 'export PATH=path_to_dnnbrain_bin:$PATH' >> ~/.bashrc
echo 'export PYTHONPATH=path_to_dnnbrain:$PYTHONPATH' >> ~/.bashrc

For tcsh/csh shell, run the following commands to set environment variables:

echo 'set DNNBRAIN_DATA=path_to_dnnbrain_data' >> ~/.cshrc
echo 'set PATH=(path_to_dnnbrain_bin $PATH)' >> ~/.cshrc
echo 'set PYTHONPATH=(path_to_dnnbrain $PYTHONPATH)' >> ~/.cshrc

Alternatively, you can also set environment variables by editing .bashrc or .cshrc file in a text editor.

You have to make these changes be effective by openning a new terminal window or “source” the .bashrc/.cshrc file in the current terminal window.

Windows¶

Note: DNNBrain scripts in path_to_dnnbrain_bin are only executable in UNIX-like shell, such as git bash.

Open the environment variable editor which can be found under “Windows Menu —> Control Panel —> Advanced System Settings —> Environment Variables”. Then:

Create a new environment variable named DNNBRAIN_DATA with value as path_to_dnnbrain_data.
Add path_to_dnnbrain_bin to PATH.
Add path_to_dnnbrain to PYTHONPATH (If PYTHONPATH is not existed, create a new one).

Separate multiple paths with semicolons (;).

Alternatively, you can use the following commands in PowerShell to complete the above 3 steps (Please replace the path between asterisks with the real path, for example, the *path_to_dnnbrain* should be replaced with the 'F:\Python3.6.5\Lib\site-packages\dnnbrain')

$new_path = *path_to_dnnbrain_data*
[environment]::SetEnvironmentvariable('DNNBRAIN_DATA', $new_path, "User")

$old_path = [environment]::GetEnvironmentvariable("PATH", "User")
$path_to_dnnbrain_data_bin = *path_to_dnnbrain_data_bin*
$new_path=$old_path,$path_to_dnnbrain_data_bin -Join ";"
[environment]::SetEnvironmentvariable("PATH", $new_path, "User")

$old_path = [environment]::GetEnvironmentvariable("PYTHONPATH", "User")
$path_to_dnnbrain = *path_to_dnnbrain*
$new_path=$old_path,$path_to_dnnbrain -Join ";"
[environment]::SetEnvironmentvariable("PYTHONPATH", $new_path, "User")

1.5. Download DNN parameters¶

The pretrained parameters are always necessary and they should be placed in the “models” directory under the DNNBRAIN_DATA.

The pretrained parameters of DNNs supported by DNNBrain are shown as below. You can download preferred DNN parameters by clicking corresponding filenames (Make sure to rename the downloaded file as its filename used here).