inference

Roboflow Inference is an open-source server and workflow engine designed to turn any computer or edge device into a command center for computer vision projects. It allows developers to self-host their own fine-tuned models, access state-of-the-art foundation models (such as Florence-2, CLIP, and SAM2), and build composable computer vision workflows using a visual interface or API. Key features include: - Self-hosted deployment on various devices (cloud servers, NVIDIA Jetson, Raspberry Pi, Mac, Windows, Linux). - A composable workflow system to chain models, track objects, count, measure, and visualize results. - Integration with traditional computer vision methods (OCR, barcode reading, template matching). - Real-time video stream processing and management (RTSP, webcam). - REST API and Python SDK (`inference-sdk`) for seamless integration into applications. - Connection to external systems and notification triggers (email, SMS, webhooks).