jldowns / vatic-docker-contrib

VATIC in a box

vatic-docker-contrib

This is a Docker build of the contrib branch of VATIC using the supported install script as much as possible.

Github: https://github.com/jldowns/vatic-docker-contrib

Dockerhub: https://hub.docker.com/r/jldowns/vatic-docker-contrib/

Run using:

docker run -it -p 8080:80 -v $DATA_DIR:/home/vagrant/vagrant_data jldowns/vatic-docker-contrib:0.1

I've tested this container for local work only, using Docker 17.03 on MacOS Sierra.

When using this software, please cite the authors:

Carl Vondrick, Donald Patterson, Deva Ramanan. "Efficiently Scaling Up
Crowdsourced Video Annotation" International Journal of Computer Vision
(IJCV). June 2012.

User Guide

Create a data folder to share with Docker:

DATA_DIR=`pwd`/data/
mkdir -p $DATA_DIR

Start the VATIC server using

docker run -it -p 8080:80 -v $DATA_DIR:/home/vagrant/vagrant_data jldowns/vatic-docker-contrib:0.1

I highly recommend using a version tag if you plan to use this container in a project. :latest will always point to the latest version, and is subject to change.

This should open up a shell inside the container. Start the MySQL and Apache services by executing

/home/vagrant/start_services.sh

Extracting Frames

Copy a video into DATA_DIR and run the following command to extract the frames into a directory called your_video:

cd /home/vagrant/vatic
turkic extract /home/vagrant/vagrant_data/your_video.mp4 /home/vagrant/vagrant_data/your_video_frames/

Note that right now the turkic command only works while in the /home/vagrant/vatic directory.

Load and Publish

The following command loads a directory of frames into VATIC, and sets the labels. job_id is a unique handle you set to reference this job.

turkic load job_id /home/vagrant/vagrant_data/your_video_frames/ car skateboard airplane --offline

VATIC automatically splits long videos into smaller pieces. Each piece gets its own unique URL. To see all the URLs you can type:

turkic publish --offline

This will return something along the lines of

http://localhost/?id=1&hitId=offline
http://localhost/?id=2&hitId=offline
http://localhost/?id=3&hitId=offline
http://localhost/?id=4&hitId=offline
http://localhost/?id=5&hitId=offline

When accessing those URLs on your host machine, make sure you include the port number. If you used the command above, you would go to http://localhost:8080/?id=1&hitId=offline to start annotating the first video.

Reading the annotations

After annotating each segment, VATIC will save the annotation in a machine readable format by typing:

turkic dump job_id -o /home/vagrant/vagrant_data/annotations.txt

If you prefer JSON, you can save the annotations as JSON with the following:

turkic dump job_id -o /home/vagrant/vagrant_data/annotations.json --json

Full syntax and features are covered in depth at https://github.com/cvondrick/vatic/tree/contrib.

FAQ

I get the error No handlers could be found for logger "turkic.geolocation"

Make sure you are in the /home/vagrant/vatic directory before running the turkic command.

I accidentally exited before dumping the annotations!

Find the container you just stopped by typing

docker ps -a

You can then restart and reattach the container and dump your data by typing

docker start $JOB
docker attach $JOB

where $JOB is the container ID.