Get started with doccano

What is doccano?

doccano is an open-source data labeling tool for machine learning practitioners. You can use doccano to perform different types of labeling tasks with many data formats. To see what doccano can do, try the doccano demo.

You can also integrate doccano with your script via the doccano REST APIs. By using the doccano APIs, you can label your data by using some machine learning model.

Doccano labeling workflow

To complete a labeling project with doccano:

Install doccano.
Run doccano.
Set up the labeling project. Select the type of labeling project and configure project settings.
Import your dataset. You can also import labeled datasets.
Add users to the project.
Define the annotation guideline.
Start labeling the data.
Export the labeled dataset.

Quickstart

Install doccano with pip (Python 3.8+):
```
pip install doccano
```

Run doccano:

doccano init
doccano createuser
doccano webserver

# In another terminal, run the command:
doccano task

Open the doccano UI at http://localhost:8000/auth.
Sign in with the username and password created by doccano createuser. The default is username: admin, password: password.
Change the default admin password at http://localhost:8000/admin/password_change/.
Return to the doccano UI at http://localhost:8000/projects?.
Create a project for labeling data. Click Create, select a project type, and fill out project details.
Import a dataset. Go to the Dataset page and click Actions > Import Dataset and import the dataset you want to use.
Click Annotate and label the data.
When you're finished, export the labeled dataset. Go to the Dataset page and click Actions > Export dataset.

Architecture

You can customize doccano to suit your needs. The architecture of doccano consists of two parts: backend and frontend.

Module	Technology	Description
doccano backend	Python, Django, and Django Rest Framework	Perform data labeling via REST APIs.
doccano frontend	Javascript web app using Vue.js and Nuxt.js	Perform data labeling in a user interface.

Contact

If you get stuck, check the FAQ.

For help and feedback, feel free to contact the author.

2.9 KiB Raw Blame History