Roboflow Inference

Roboflow Inference runs computer vision models on images and camera feeds for teams building inspection, monitoring and image analysis systems. You can host it on your own computer, server or edge device.

• Models and workflows: Combine detection, classification and segmentation with OCR, barcode reading, QR reading and template matching, using models such as Florence-2, CLIP and SAM2.
• Tracking and measurement: Build workflows that count objects, track movement, measure time in a zone or handle tasks such as license plate reading and face blurring.
• Live video: Process RTSP streams and webcams with hardware acceleration and GPU batching, and record predictions for later analysis.
• Application connections: Use the Python SDK or REST API to connect results to other software, trigger email or Twilio notifications, or call webhooks; custom code and models can extend workflows.

Docker deployments support Linux, Windows and macOS, with support for NVIDIA Jetson and Raspberry Pi devices. NVIDIA CUDA provides GPU acceleration where available. Processing on your own hardware is free; optional hosted servers and the Serverless Cloud API process data on Roboflow infrastructure.

Pre-trained models, foundation models and public workflows don't require an API key. Fine-tuned models and private workflows require a Roboflow API key. Remote stream management through the Roboflow UI requires an account and linked key.