Inference
A self-hostable inference server that turns any computer or edge device into a command center for computer vision: running fine-tuned or foundation vision models, chaining them into workflows, and connecting cameras, video streams and external systems.
Inference turns a computer or edge device into a server for computer vision. It runs self-hosted, fine-tuned models as well as foundation vision models such as Florence-2, CLIP and SAM2, and supports object detection, classification and instance segmentation models including several YOLO versions, alongside traditional computer-vision methods such as OCR, barcode and QR reading and template matching.
Workflows are composable blocks that give different models a common interface, so they can be chained together, combined with large multimodal models for higher-level decisions, and extended with custom code to track, count, time, measure and visualize objects, connect to external systems and APIs, and send notifications when an event happens, turning simple model predictions into a computer-vision microservice or a self-contained visual agent running on a video stream.
It installs through a Python package and runs as a Docker container, with GPU acceleration through the NVIDIA Container Toolkit, and includes a development mode with a bundled Jupyter notebook quickstart and a UI for building and deploying Workflows.