inference
Turn any computer or edge device into a command center for your computer vision projects.
What is inference?
Turn any computer or edge device into a command center for your computer vision projects.
Overview of inference
What it is
Inference is an open-source tool that turns any computer or edge device into a command center for computer vision projects. It allows users to self-host fine-tuned models, access foundation models like Florence-2, CLIP, and SAM2, and build workflows that combine machine learning with traditional computer vision methods such as OCR, barcode reading, QR code scanning, and template matching. The project is maintained by Roboflow and is available on GitHub under an open-source license.
Key features
- Self-host your own fine-tuned models
- Access foundation models (Florence-2, CLIP, SAM2)
- Use Workflows to track, count, time, measure, and visualize
- Combine ML with traditional CV methods (OCR, Barcode Reading, QR, template matching)
- Monitor, record, and analyze predictions
- Manage cameras and video streams
- Send notifications when events happen and connect with external systems and APIs
Use cases
Inference is suitable for building computer vision applications that require real-time analysis of video streams, such as detecting small objects with SAHI, multi-model consensus, active learning, reading license plates, blurring faces, and background removal. It can be used to create smart parking systems, self-serve checkout systems, and other visual agents that run on live video streams.
Getting started
Install Docker (and NVIDIA Container Toolkit for GPU acceleration if you have a CUDA-enabled GPU). Then run:pip install inference-cli && inference server start --devThis will pull the proper image for your machine and start it in development mode. A Jupyter notebook server with a quickstart guide runs on http://localhost:9001/notebook/start.
License and pricing
The license is not stated in the provided metadata (NOASSERTION). The project is open-source. Pricing for hosted compute (Serverless Cloud API and Dedicated Deployments) is available on the Roboflow pricing page.
Based on the project's public README and repository metadata.