supervisely/supervisely

Supervisely – a web‑based platform for computer‑vision projects

What it is – Supervisely is a cloud‑hosted “operating system” for computer‑vision work. It bundles data labeling (images, video, 3‑D point clouds, medical DICOM), data‑visualisation, model training, inference, quality‑control, synthetic‑data generation and collaboration tools into a single web UI. The platform is extensible through a public Python SDK and a rich REST API, letting you write scripts or full‑blown apps that run inside the Supervisely ecosystem.

Key components

• Supervisely Platform – the core web service (available at https://app.supervisely.com) that stores projects, datasets, annotations and models.
• Supervisely Ecosystem – a marketplace of hundreds of ready‑to‑run apps (head‑less scripts, interactive UI tools, labeling‑tool extensions) that can be launched with one click.
• Python SDK (supervisely package) – thin wrapper around the REST API that handles authentication, pagination, error handling, etc. Example workflow:

  import supervisely as sly
  api = sly.Api.from_env()
  project = api.project.create(workspace_id=123, name="demo project")
  dataset = api.dataset.create(project.id, "dataset-01")
  img = api.image.upload_path(dataset.id, "img.png", "/local/path/img.png")
  ann = api.annotation.upload_path(img.id, "/local/path/ann.json")
  

• App framework – any web server (FastAPI, Flask, etc.) can be turned into a Supervisely app. The framework supplies UI widgets, authentication, versioning and deployment hooks, so developers can focus on the computer‑vision logic.

How you can extend it

Level	What you build	Typical use‑case
1️⃣	Direct HTTP calls	Quick automation from any language
2️⃣	Python scripts using the SDK	Data import/export, batch model inference
3️⃣	Headless Python apps (no UI)	Custom format converters, bulk user management
4️⃣	Apps with interactive UIs	Custom labeling interfaces, training dashboards
5️⃣	UI‑integrated labeling tools	AI‑assisted annotation directly inside the labeling UI

Main selling points

• Start in a minute – pip install supervisely gets you the SDK; the web UI is ready instantly.
• Magically simple API – the SDK abstracts the REST endpoints, handling retries, validation and pagination.
• Customizable everywhere – you can replace UI widgets, plug in your own models, or write private apps.
• One‑click deployment – apps are packaged as Docker images (e.g., supervisely/agent) and can be launched from the platform UI.
• Versioned releases – the ecosystem supports branches and releases on both GitHub and GitLab.
• Community & enterprise – used by >65 000 users, including Fortune‑500 companies; apps can be public or private.

Typical users

• Data scientists who need a managed environment for labeling, training and evaluating CV models.
• Teams that require collaborative annotation (images, video, 3‑D, medical).
• Engineers building custom pipelines (e.g., synthetic data generation, model‑as‑service) that must integrate with existing Supervisely projects.
• Companies looking for an extensible platform rather than a closed‑source product.

Installation / quick start

pip install supervisely   # install the Python SDK
# set environment variable SUPERVISELY_API_TOKEN with your token
python -c "import supervisely as sly; api = sly.Api.from_env(); print(api.team.get_my_team())"

For full apps, you typically write a FastAPI service, add UI widgets via the SDK, and push the Docker image to the platform.

Resources

• Website: https://supervisely.com
• SDK docs: https://developer.supervisely.com
• API reference: https://api.docs.supervisely.com
• Ecosystem of apps: https://ecosystem.supervisely.com
• YouTube course: What is Supervisely?

---

All information above is taken directly from the repository’s README; no external assumptions have been added.