Skip to content

Latest commit

 

History

History
115 lines (84 loc) · 5.69 KB

installation.md

File metadata and controls

115 lines (84 loc) · 5.69 KB

SimBAxTF or SimBA?

!!! IMPORTANT !!! You can choose to install SimBA as a standalone package or install SimBA with TensorFlow integration.

  1. If you would like to be able to call DeepLabCut or DeepPoseKit commands via the SimBA interface (whuch requires a local GPU), please install SimBAxTF from the master branch. Please see the SimBA tutorials on DeepLabCut or DeepPoseKit for information on what it means to run DeepLabCut and DeepPoseKit within the SimBA GUI. See full installation instructions below.

  2. If you do not want to use TensorFlow through SimBA on your local machine, and instead have DeepLabCut/DeepPoseKit/SLEAP installed elsewhere, please install SimBA from the SimBA_no_TF branch. This does not require a GPU, or local installations of DeepLabCut, DeepPoseKit, or SLEAP. Please see full instructions below. This non-TF version of SimBA includes all functionalities of SimBAxTF, except for the ability to generate pose-estimation models through the SimBA GUI. Pose-estmation model results can still be imported and analysed.

  3. If you are on a Linux machine (or Mac or Windows PC), or you just want to work with a far speedier (and buggier!!) version of SimBA then run the SimBAxTF-dev version. This version is the same as the SimBAxTF, but contains the latest and the greatest tools (most undocumented, as yet) for explainable and interpretable supervised models for behavioral neuroscience.

Requirements

  1. Python 3.6 <-- VALIDATED WITH 3.6.0
  2. Git
  3. FFmpeg
  4. Microsoft Windows operating system

Installing SimBA Option 1 (RECOMMENDED!)

Install SimBAxTF with integrated TensorFlow (use this installation method when running DeepLabCut, DeepPoseKit, or SLEAP locally using a GPU)

Open bash or command prompt and run the following commands on current working directory

pip install simba-uw-tf

Install SimBA standalone package (without TensorFlow or integrated DeepLabCut/DeepPoseKit support)

Open bash or command prompt and run the following commands on current working directory

pip install simba-uw-no-tf

Install SimBAxTF-development version

Open bash or command prompt and run the following commands on current working directory

pip install simba-uw-tf-dev

How to launch SimBA ( installed using pip install simba-uw-tf)

  1. Open up command prompt anywhere.

  2. In the command prompt type

simba
  1. Hit Enter.

Note: If you installed SimBA on a virtual environment (anaconda), after installation, you may have to run run conda install shapely for SimBA to work.

Installing SimBA Option 2

Install SimBAxTF with integrated TensorFlow (use this installation method when running DeepLabCut or DeepPoseKit locally using a GPU)

Open bash or command prompt and run the following commands on current working directory

git clone -b master https://github.com/sgoldenlab/simba.git

pip3 install -r simba/simba/requirements.txt

Install SimBA standalone package (without TensorFlow or integrated DeepLabCut/DeepPoseKit support)

Open bash or command prompt and run the following commands on current working directory

git clone -b SimBA_no_TF https://github.com/sgoldenlab/simba.git

pip3 install -r simba/SimBA/requirements.txt

How to launch SimBA (installing by cloning)

  1. Open up command prompt in the SimBA folder

  2. In the command prompt type

python SimBA.py
  1. Hit Enter.

Note: For this launch to work you need to add python to the environmental path.

python dependencies

package ver.
Pillow 5.4.1
deeplabcut 2.0.9
eli5 0.10.1
imblearn 0.5.0
imutils 0.5.2
matplotlib 3.0.3
Shapely 1.6.4.post2
deepposekit 0.3.5
dtreeviz 0.8.1
opencv_python 3.4.5.20
numpy 1.18.1
imgaug 0.4.0
pandas 0.25.3
scikit_image 0.14.2
scipy 1.1.0
seaborn 0.9.0
sklearn 1.1.0
scikit-learn 0.22.1
tensorflow_gpu 0.14.1
scikit-learn 0.22.1
tqdm 4.30.0
yellowbrick 0.9.1
xgboost 0.9
tabulate 0.8.3
tables ≥ 3.5.1