Docker Environment for DeepLabCut

This package will allow you to run DeepLabCut with everything pre-installed inside a Docker container.

This Docker file is based off the Bethge lab container. Specifically, the one we provide comes with Ubuntu 14.04 + Cuda 8.0 + CuDNN v5 and Tensorflow 1.2 (ideal for the current version of DeepLabCut), and the required python packages.

NOTE: this container does not work on windows hosts!

Prerequisites

(1) Install Docker. See https://docs.docker.com/install/ & for Ubuntu: https://docs.docker.com/install/linux/docker-ce/ubuntu/

(2) Install nvidia-docker, see https://github.com/nvidia/nvidia-docker/wiki/Installation-(version-1.0). Specifically for linux follow, https://nvidia.github.io/nvidia-docker/. But, basically it is just:

$ sudo apt-get install nvidia-docker

TIP: You can test your nvidia-docker installation by running:

nvidia-docker run --rm nvidia/cuda nvidia-smi

(3) Add your user to the docker group (https://docs.docker.com/install/linux/linux-postinstall/#manage-docker-as-a-non-root-user) Quick guide to create the docker group and add your user: Create the docker group.

$ sudo groupadd docker

Add your user to the docker group.

$ sudo usermod -aG docker $USER

(perhaps open a new terminal to make sure that you are added from now on) Lastly, install the Docker and DeepLabCut:

git clone https://github.com/AlexEMG/Docker4DeepLabCut
git clone https://github.com/AlexEMG/DeepLabCut
cd Docker4DeepLabCut

Step-by-step instructions for creating the Docker environment

Next create the image. The image needs only be created once. All the required software will be downloaded from DockerHub. (you can pick a user name and container name other than dlc_user/dlc_tf1.2 if you want):

docker image build -t dlc_user/dlc_tf1.2 .

Starting the docker container from your image (de-novo):

In the terminal start your container with the following command (in the DockerContainer4DeepLabCut folder):

Options:

• change port: (i.e. 2351 can be 777, etc)
• change which GPU to use (check which GPU you want to use in the terminal by running "nvidia-smi")
• change the name: --name alex_GPU1 (i.e. alex_GPU1 can be anything you want)
• change the home fodler:-e USER_HOME=$HOME/DeepLabCut (i.e. this can be -e USER_HOME=$HOME/whateveryouwant)

GPU=1 bash ./dlc-docker run -d -p 2351:8888 -e USER_HOME=$HOME/DeepLabCut --name alex_GPU1 dlc_user/dlc_tf1.2

Do not run this with sudo. Now you can enter your container in the terminal, or via a browser interface:

• Enter the container via the terminal (to get terminal access in container):

docker exec -it alex_GPU1 /bin/bash

Access your linked (internal home) directory:

cd ../../../home/

• Enter the container via a browser interface (i.e Jupyter Notebook):

Go to the port you specified, i.e. in our example enter http://localhost:2351 in Google Chrome. Get the token: in the terminal look at the docker log; copy and paste the value after "token=":

$ docker logs alex_GPU1 

Now you have an Ubuntu with Python3 and a GPU-installed with Tensorflow 1.2, and all the other dependencies ... installed! Happy DeepLabCutting!

Docker Quick Tips:

Check which containers are running:

$ docker ps 

You can stop a container:

$ docker stop alex_GPU1 

You can re-start your container:

$ docker start alex_GPU1

After stopping you can remove old containers:

$ docker rm alex_GPU1

(once removed, it can be created again):

GPU=1 bash ./dlc-docker run -d -p 2351:8888 -e USER_HOME=$HOME/DeepLabCut --name alex_GPU1 dlc_user/dlc_tf1.2

DeepLabCut Quick Training Guide & Evaluation Tips can be found on the DeepLabCut Wiki