NVIDIA/DeepStream

NVIDIA/DeepStream

Description: DeepStream OSS

Language: C++

License: NOASSERTION

Stars: 5

Forks: 0

Open issues: 0

Created: 2026-05-08T03:41:15Z

Pushed: 2026-06-03T09:46:34Z

Default branch: main

Fork: no

Archived: no

README:

DeepStream

NVIDIA DeepStream SDK is a streaming analytics toolkit for AI-based video and image understanding, providing a GStreamer-based framework to build multi-stream, multi-model inference pipelines on NVIDIA GPUs (dGPU and Jetson).

DeepStream pipelines combine hardware-accelerated decoding/encoding, TensorRT inference, object tracking, and message-broker integrations to deliver real-time video analytics across dGPU and Jetson platforms.

Overview

This repository contains the complete source code for DeepStream 9.0.

Components ([src/](src/README.md)):

• [src/gst-plugins/](src/README.md#gstreamer-plugins) — DeepStream GStreamer plugin sources
• [src/utils/](src/README.md#utility-libraries) — utility library sources
• [src/apps/sample_apps/](src/README.md#sample-applications) — GStreamer-based sample applications
• [src/apps/reference_apps/](src/README.md#reference-applications) — advanced reference applications
• [src/apps/tao_apps/](src/README.md#tao-apps) — TAO-model integration apps
• [src/service-maker/](src/README.md#service-maker) — Service Maker C++/Python SDK and apps

Tools ([tools/](tools/)):

• `inference_builder` — visual inference pipeline builder
• `auto-magic-calib` — camera auto-calibration tool
• [yolo_deepstream](tools/yolo_deepstream/README.md) — YOLO + TensorRT integration
• [sam2-onnx-tensorrt](tools/sam2-onnx-tensorrt/README.md) — SAM2 ONNX-to-TensorRT conversion

AI agent skills ([skills/](skills/README.md), for Claude Code & compatible coding agents):

• [deepstream-dev](skills/deepstream-dev/SKILL.md) — general DeepStream development
• [deepstream-import-vision-model](skills/deepstream-import-vision-model/SKILL.md) — autonomous vision-model onboarding

Requirements

Before building, ensure the following prerequisites are installed:

• NVIDIA compute stack — driver, CUDA, cuDNN, and TensorRT at the versions listed below. See the DeepStream SDK Installation Guide.
• DeepStream 9.0 — installed via the DS 9.0 public Debian package and its install.sh.

> SBSA / DGX Spark: no DS deb package exists for this platform — use the NVIDIA SBSA Docker container, which bundles the compute stack and DeepStream. See [build/BUILD.md](build/BUILD.md).

Getting Started

# Install Git LFS (required for sample video streams used by some apps)
sudo apt-get install git-lfs

# Clone the repo with submodules
git clone --recurse-submodules https://github.com/NVIDIA/DeepStream.git && cd DeepStream

# Pull LFS-tracked files
git lfs install && git lfs pull

See [build/BUILD.md](build/BUILD.md) for full build instructions, including system package dependencies (x86, aarch64, SBSA / DGX Spark), build/build.sh usage and environment variables (CUDA_VER, NVDS_VERSION), and build output locations under /opt/nvidia/deepstream/deepstream-9.0/.

Supported Platforms

| Platform | Architecture | Notes | |---|---|---| | x86 dGPU | x86_64 | Ubuntu 24.04, CUDA 13.1, TensorRT 10.14.x, driver 590+ | | Jetson | aarch64 | JetPack 7.1 GA (CUDA 13.0, TensorRT 10.13.x) | | SBSA / DGX Spark | aarch64 | No DS tar/deb package; build and run inside the NVIDIA SBSA Docker container |

# Build. The script prompts for sudo only when installing to system paths.
bash build/build.sh

Usage

After bash build/build.sh, binaries are installed to /opt/nvidia/deepstream/deepstream-9.0/bin/. Run the reference deepstream-app with one of the sample configs:

cd /opt/nvidia/deepstream/deepstream-9.0/samples/configs/deepstream-app
deepstream-app -c source30_1080p_dec_infer-resnet_tiled_display.txt

# After the first install, clear the GStreamer plugin cache if needed:
rm -rf ~/.cache/gstreamer-1.0/

• More such examples: DeepStream Quickstart Guide
• Detailed API reference: DeepStream SDK Developer Guide

Each app must be run from its source directory so relative config paths resolve correctly. Refer to the README inside each app directory for app-specific run instructions and config options.

Running with Triton Inference Server (Docker)

> Optional. Skip this if the bare-metal build above already works for you. Use the Triton container when you need Triton-backed inference, or when you're on SBSA / DGX Spark (where no native DS deb package exists).

NVIDIA publishes a DeepStream Docker image bundled with Triton Inference Server.

# One-time NGC login (get an API key from https://ngc.nvidia.com)
docker login nvcr.io # username: $oauthtoken, password:

# Pull the image (use 9.0-triton-arm-sbsa instead on SBSA / DGX Spark)
docker pull nvcr.io/nvidia/deepstream:9.0-triton-multiarch

# Launch with display (use 'fakesink' in your pipeline for headless)
export DISPLAY=:0 && xhost +
docker run -it --rm --gpus all --network=host \
-e DISPLAY=$DISPLAY -v /tmp/.X11-unix/:/tmp/.X11-unix \
nvcr.io/nvidia/deepstream:9.0-triton-multiarch

# Inside the container, run a Triton-backed sample
cd /opt/nvidia/deepstream/deepstream-9.0/samples/configs/deepstream-app-triton
deepstream-app -c source30_1080p_dec_infer-resnet_tiled_display.txt

Prerequisites: Docker (docker-ce), the NVIDIA Container Toolkit, NVIDIA driver 590+, and an NGC API key. Triton sample model repos ship under /opt/nvidia/deepstream/deepstream/samples/triton_model_repo/. For gRPC-backed Triton, use samples/configs/deepstream-app-triton-grpc/ instead.

Documentation

| Page | Description | |------|-------------| | [Overview &…