Delfin is a Multi-Task Annotation Tool with Active Learning Ambitions

Documentation: WIP

---

Delfin

As the need for annotated data is becoming apparent, Delfin aims to provide a unified application to manage and accelerate generation of labeled data.

Some of it's features include:

• Multi-Task framework: treat each dataset as an opportunity to extract multiple scopes of labels (e.g.
• Built for Active Learning: Delfin aims to provide a convenient interface for Machine Learning practitioners to embed active learning and semi-supervised learning approaches into their annotation workflow.
• Extensible: both the backend and frontend is built to be easily extensible with new annotation types and data modalities.
• Easily Deployable: dockerized and ready to deploy on your favourite cloud or on prem servers.

Built On

• Python 3.7+
• Vue.js

Initialize Environment (Outside of Docker)

export PYTHONPATH="${PYTHONPATH}:<path-to-repo>/delfin/backend/src/"
set -o allexport; source .env.local; set +o allexport;

Deploy