Introduction

Vision SDK

Vision SDK turns a single config file into a production video pipeline - inputs, models, tracking, and outputs - running on NVIDIA DeepStream.

Define your input streams, AI models, object tracker, and outputs in a single YAML config file, then run everything with one command.

Vision SDK pipeline overview - inputs (live streams, video files) flow through models (detector, tracker, classifier) into outputs (display, file recording, RTSP/RTMP streams, Kafka, FPS counter), all configured via a single YAML file.

Where to next

Prerequisites - what you need installed
Installation - drop the wheel into DeepStream
Quick start - your first pipeline in three steps