GeezerVison

The World Wide Web, restored to 800 pixels of glory.

GeezerVison is showing docs.roboflow.com — Python SDK | Reference | Roboflow Docs

For the complete documentation index, see llms.txt. This page is also available as Markdown.

The Roboflow Python SDK is a wrapper around the Roboflow REST API, providing abstract methods for interacting with Roboflow in Python code.

The Python SDK offers methods for managing projects and workspaces, uploading and downloading datasets, running inference on models, uploading model weights, and more.

Examples and references for using methods in the Python SDK are documented in the API Reference documentation.

Resources

Installation

The Roboflow Python package requires Python >=3.10.

pip install roboflow

Lite Package

If you only need the Vision Events SDK and the CLI without heavy image processing dependencies (OpenCV, NumPy, Pillow), install the lightweight variant instead:

pip install roboflow-slim

The roboflow-slim package is significantly smaller than the full roboflow package because it omits NumPy, OpenCV, Matplotlib, and Pillow. It is well-suited for edge devices, serverless environments, CI/CD pipelines, and microservices that only need vision-event logging or basic CLI operations.

Both packages share the same codebase and API. Code written against roboflow-slim works identically against the full roboflow package - drop the lighter dependency where you don't need image visualization.

Your first script

Grab your API key from Settings → API Key and export it:

Then download a version of a public Universe dataset:

See Authenticate with the Python SDK for the other two ways to pass a key. To run inference, use a Workflow or the REST API - SDK inference is deprecated.

Quick Reference

The SDK's main objects:

Class

What it represents

Common methods

Roboflow

Top-level client

workspace(), project()

Workspace

A Roboflow workspace

list_projects(), create_project(), upload_dataset(), list_workflows(), list_folders(), trash()

Project

A dataset/project

versions(), version(n), upload(), train(), delete(), restore()

Version

A frozen dataset version

download(), train(), model, delete(), restore()

*.Model (per task)

A hosted trained model

predict()

For task-oriented walkthroughs of every public method, browse the rest of this section in the sidebar.

Microsoft Visual C++ Instructions

Microsoft Visual C++ redistributable, is needed for some Python packages. You can download it from the official Microsoft website:

For x64 systems: https://aka.ms/vs/16/release/vc_redist.x64.exe

For x86 systems: https://aka.ms/vs/16/release/vc_redist.x86.exe

After downloading the appropriate file, run the installer and follow the prompts to install the redistributable.

PreviousInstall and Set Up the CLI

NextAuthenticate with the Python SDK

Last updated 3 months ago

Was this helpful?

You are visitor 01402503 — GeezerVison does not store the pages you look up.
Best viewed with Netscape Navigator 4.7 or Internet Explorer 6 at 800×600. No cookies required.