Ezlabels
Ezlabels is an Open Source tool that solves the task of labeling a custom dataset quickly, efficiently and privately. It is designed for installation on your local machine or your own servers and to be accessed by multiple users simultaneously within one or more markup projects.
The basic flow that is implemented right now is built around the task of tagging an image dataset (you can define desired classes) and accepting or rejecting images (binary classification). We plan on making this process easily extendable in Rails fashion, with custom engines and gems.
Disclaimer
This project is at a very early stage (in fact, it was assembled for a one-off task and then forgotten) and most likely you will struggle to make use of it ¯_(ツ)_/¯ But PRs are most welcome!
What's implemented and what's planned
- Multiple users
- Multiple projects
- Multiple users can work on the same project (each one works on their own part)
- Deploy process
- Image tagging
- Image binary classification
- Hotkeys
- Simple charts
- Export/import as CSV (currently you can only manipulate data via
psqlorrails console) (#9 and #8) - Engines for text and video labeling, object detection and background segmentation
- REST API for bots and pluggable services (#20)
- Data comments (#19)
- Data flows (i.e. take a subset of an existing dataset and create an independent project with it) (#18)
- S3/AWS integration (#10)
- Role model and project visibility
Prerequisites
You should have Ruby and PostgreSQL installed locally. The easiest way to do this depends on your OS and your taste.
The recommended way to deal with Ruby is via rbenv. Don't forget to install ruby-build.
This will show you the required ruby version:
cat .ruby-version
PostgreSQL: macOS
The easiest way to install is via Postgres App.
PostgreSQL: Linux
There are more choices for Linux, but this one usually works well:
https://www.digitalocean.com/community/tutorials/how-to-install-and-use-postgresql-on-ubuntu-16-04
Configuration
The app follows the https://12factor.net/ guidelines. All configuration can be done via env variables (dotenv is responsible for that).
First, copy the configuration file sample:
cp .env.example .env
Second, edit it with your editor of choice:
nano .env
The variable names are self-descriptive, but you are especially interested in DATABASE_URL (more on how it works you can find in this section of Rails Guidelines).
Environment Variables
| Variable Name | Description |
|---|---|
| DATABASE_URL | postgres://postgres:postgres@localhost/ezlabels_development |
| CAP_APP_NAME | App name one the server (i.e. folder name at /var/www/) |
| CAP_REPO_URL | Where to pull the repo from to the server |
| DEVISE_SECRET | Salt for Devise ruby gem (your user passwords will be created with it) |
Basic setup
bundle install
bundle exec rake db:setup
TODO: Provisioning
This will do for now:
We're waiting for someone smart to submit a Pull Request with Docker or Ansible workflow.
Deploy
Deploy is done via Capistrano. If you're a fan of Heroku, it is probably going to work as is.
Configuration
Change CAP_APP_NAME and CAP_REPO_URL in .env to those you want to use.
Then, setup the remote directory structure:
bundle exec cap production deploy initial
bundle exec cap production puma:nginx_config
Deploy
bundle exec cap production deploy