Skip to content

Introduction to Roboflow Ecosystem

Roboflow Inference banner

Roboflow provides everything you need to label, train, and deploy computer vision solutions. It helps you manage and refine datasets, provides tools to streamline and speed up data labelling, helps train and deploy models in the cloud and on edge devices.

Inference is what allows you to deploy and run computer vision models. It enables you to perform object detection, classification, instance segmentation and keypoint detection, and utilize foundation models like CLIP, Segment Anything, and YOLO-World, through a Python-native package, a self-hosted inference server, or a fully managed API.

Roboflow offers both a free tier, and paid plans, encompassing all of its products.

  • Over half of Fortune 100 companies build with Roboflow. If you're an enterprise customer and interested in a custom solution, parallel processing, active learning or licenses for more than one cloud instance or edge device - Reach Out to our sales team!

Inference is commonly used with several other roboflow products.

Roboflow App

Roboflow App This is your central dashboard. Here you can upload data, annotate images, define datasets, train and deploy models. You can find the API key (scoped to workspace)

roboflow Package

Your workspace can be managed via the dashboard UI. If you'd like to do it via Python, install the roboflow package. This lets you manage your workspace, upload datasets and model weights, and even run model inference (a bit outdated).

However, If all you need is to run a deployed model, you likely won't need roboflow at all. Where possible, we recommend inference. That's what we use on our servers!

If you wish to use the roboflow package, instructions can be found in Roboflow Python Package Docs.


Universe is our space for sharing datasets and models.

Search for models, test out their performance on your images, track model versions, access in inference, build on top.

The model_id you pass into inference can be a model_id from Universe.


What happens when you infer some results from a model? supervision lets you plot bounding boxes and segmentation masks, track objects, merge various detections. With supervision tools such as InferenceSlicer you can detect small objects in an image by running the model on smaller patches.

It also encodes model results from various sources - inference, Hugging Face, Ultralytics and more, into a common format.


Workflows are an new inference feature. Instead of writing code, you may chain together blocks to build your computer vision algorithms from scratch. There's an expanding library of blocks available - see if you find anything you like!