roboflow/roboflow-python
The official Roboflow Python package. Manage your datasets, models, and deployments. Roboflow has everything you need to build a computer vision application.
What it solves
It provides a programmatic way to build, manage, and deploy computer vision models. It eliminates the need to manually interact with the Roboflow web interface for tasks like dataset management, model training, and inference execution.
How it works
The package uses a hierarchical structure (Workspace $\rightarrow$ Project $\rightarrow$ Version) to mirror the Roboflow platform. Users authenticate via API keys or a login command and can then use Python methods to upload images and annotations, create project versions, trigger training, and run predictions on hosted or self-hosted models.
Who it’s for
Computer vision developers and ML engineers who want to automate their CV pipeline, manage datasets programmatically, or integrate Roboflow's model hosting and inference capabilities into their own applications.
Highlights
- Comprehensive CV Lifecycle: Supports the entire flow from project creation and data upload to training and deployment.
- Flexible Inference: Run predictions using models hosted on Roboflow or self-hosted via Roboflow Inference.
- Dataset Export: Search and export images across workspaces in various formats like COCO, YOLOv8, and Pascal VOC.
- Lightweight Option: Offers a
roboflow-sliminstallation for embedded devices or serverless environments to reduce dependency overhead.