Doccano

Doccano is an open-source annotation tool for machine-learning practitioners who want to manage labeling projects themselves. It is a strong fit for teams that value deployment and storage choices; the SQLite upgrade warning calls for care with the default setup.

Overview

Doccano takes a dataset from import through project setup and collaborative annotation to export. Its supported text tasks include classification, sequence labeling, and sequence-to-sequence annotation, with uses such as sentiment analysis, named entity recognition, and text summarization. It also supports image and audio/video annotation, model-assisted labeling, and review workflows.

The project dates to 2018 and is open source. Rather than locking teams into a single deployment path, Doccano can be installed locally, run in containers, built from source, or deployed in the cloud. That range suits practitioners willing to make infrastructure choices; it is less compelling for teams that want a managed service without setup decisions. For more options, see Data Labeling Software.

Key features

End-to-end project workflow: Configure a project, import datasets, add users, annotate records, and export labeled data. Keeping those steps together is useful for teams running their own labeling process.

Collaborative annotation and review: Multiple people can work on annotation, and a review workflow is supported. This makes Doccano suitable for shared projects, though the facts do not establish specific review controls or team-size limits.

REST API and model assistance: Scripts can integrate through the REST API, including labeling workflows that use machine-learning models. The auto-labeling guide demonstrates Amazon Comprehend Sentiment Analysis and permits a custom REST API, giving teams a route to connect external labeling logic.

Flexible storage: Imported datasets can be stored in Amazon S3 or Google Cloud Storage. SQLite 3 is the default database, while PostgreSQL and MySQL are also described as options.

Interface touches: The project lists multi-language support, mobile support, emoji support, and a dark theme. These broaden interface options without changing the central infrastructure work.

Login integrations: Social login via GitHub and Active Directory is described, with Okta setup instructions.

Pricing

Doccano has one free, open-source plan: free. It includes the annotation tool and installation with pip, Docker, or Docker Compose. There are no paid tiers in the plan information, so readers do not need to weigh feature gates or seat quotas against a subscription; they do need to provide and manage their own installation environment. No trial or renewal term is part of this free plan.

Platforms

Doccano is available for Linux, Windows, and macOS machines running Python 3.8 or later, and has a web interface and API. Installation options include pip, Docker, Docker Compose, source, and cloud deployment. This is useful for teams choosing their own operating environment, but it makes deployment a decision the team must own. SQLite is the default, and the guide warns that upgrading while using SQLite 3 can result in database loss. Teams using that default should protect their data before upgrades or consider another supported database.

Who it's for

Doccano fits machine-learning practitioners and teams preparing labeled datasets who need a collaborative workflow, script integration, and control over deployment or storage. Its listed applications cover sentiment analysis, named entity recognition, and text summarization. It is a less natural choice for readers who want a hosted service with no infrastructure responsibilities, or who cannot accommodate the SQLite upgrade risk and database planning.

Pros and cons

Pro — Broad deployment choice: pip, containers, source, and cloud options let teams align installation with their own environment.

Pro — Useful integration and storage options: A REST API, model-assisted labeling paths, and S3 or Google Cloud Storage support give teams ways to fit annotation into surrounding workflows.

Pro — Free and open source: The plan has no paid tier or stated seat quota, making it accessible to teams prepared to operate it themselves.

Con — SQLite upgrade hazard: The installation guide warns that upgrades can lose the database when SQLite 3 is in use, so the default requires deliberate backup and upgrade practice.

Con — Self-managed deployment: The breadth of installation choices also leaves environment and database decisions to the user, which may not suit teams seeking a managed service.

Alternatives

LightlyStudio is another free option with an open-source version under Apache License 2.0; consider it if that licensing detail is central to your choice. Label Studio is a freemium alternative with a free plan and free trial. Potato may suit readers seeking a self-hosted free tool with all features included and no usage limits or paid tiers.

Amazon SageMaker Autopilot is a paid, pay-as-you-go alternative with a free trial and no minimum fees or upfront commitments for on-demand pricing; choose it if that pricing model is preferable. Datasaur offers a free plan capped at one user, 5,000 labels a year, and 100MB of storage, with a personal workspace and a Growth trial of up to 14 days; it may suit an individual who wants those explicit limits. MakeSense.ai is a free, browser-based photo-labeling tool, a narrower alternative for that task.

Prodigy is a paid alternative with a personal lifetime license priced at $390 as a one-time purchase, including 12 months of free upgrades; consider it if that license model fits. Roboflow has a freemium plan with 10 monthly credits, described as enough to train about 30 models or run 80,000 inferences, and may fit readers whose needs align with that allowance.

Verdict

Choose Doccano if you want a free, open-source labeling workflow with collaboration, API integration, and freedom to choose deployment and storage. Its main advantage is that operational flexibility; its main caution is the stated SQLite database-loss risk during upgrades. If managing installation and database safety is not acceptable, look elsewhere.