DeepLabCut

DeepLabCut is an open source Python package for markerless pose estimation of animals performing various tasks.


Versionmodulename
2.2.0.6anaconda3/2020.07 + conda/dlc

Installing DeepLabCut in your directories

DeepLabCut is an efficient method for 2D and 3D markerless pose estimation based on transfer learning with deep neural networks that achieves excellent results (i.e. you can match human labeling accuracy) with minimal training data (typically 50-200 frames). We have noticed that during the training phase, DeepLabCut tries to write data to system directories that is not permitted. For this reason, we recommend users to install DeepLabCut in their home directories. Please execute the steps as follows so that we can create an Open OnDemand Virtual Desktop application that will launch DeepLabCut for all users.


Load the anaconda3 module to put conda in your path
module load anaconda3
Create a directory in the <PI's userid>_proj folder
mkdir -p ${HOME}/<pi's username>_proj/${USER}/condaenv
Create a symlink to the condaenv directory in your home directory
ln -s ${HOME}/<pi's username>_proj/${USER}/condaenv ${HOME}/condaenv
Create the conda environment
conda env create --prefix ${HOME}/condaenv/dlc -f /share/Apps/usr/etc/dlc.yml


Open OnDemand Application

 Select DeepLabCut from the Interactive Apps Dropdown

 Choose the resources you want to use. Click "Launch" to submit your request to the SLURM scheduler.

 When resources are available, a blue "Launch DeepLabCut" button will appear. Click the button to launch a Virtual Desktop for launching DeepLabCut.

 Virtual Desktop with shortcut for launching DeepLabCut GUI and JupyterLab

 DeepLabCut GUI

 JupyterLab session

 Use my own dlc environment
Unload conda/dlc module
module unload conda/dlc
activate my own conda environment for DeepLabCut
conda activate ${HOME}/condaenv/dlc
Start DeepLabCut GUI (or JupyterLab)
python -m deeplabcut
# OR
jupyter lab

For more information, visit http://www.mackenziemathislab.org/deeplabcut