[!NOTE] The main branch is under active development and may be unstable. For production use, please pick a stable version from the Releases page.

Deep Learning Streamer (DL Streamer)

Hardware-accelerated video analytics pipelines — CPU, GPU and NPU, from a single line of code to production-grade edge AI

Get Started • Run Your Pipeline • Samples • Elements • Documentation • Contributing

What is DL Streamer?

DL Streamer is an open-source media analytics framework built on GStreamer. It lets you build video and audio intelligence pipelines — from a simple object detection command line to a multi-stream, multi-sensor production deployment, all with minimal code.

• Powered by OpenVINO™ for optimized inference on Intel CPU, GPU, and NPU.
• Pipelines are described as simple strings (or Python/C++ code) and executed with full hardware acceleration.
• Ships with 30+ ready-to-run samples covering detection, classification, tracking, VLMs, LiDAR and more.
• Part of the Intel Open Edge Platform.

Why DL Streamer?

Quick Start - Installation

Step 1 — Install GPU/NPU drivers (required for Docker and native install)

cd ~
wget https://raw.githubusercontent.com/open-edge-platform/dlstreamer/main/scripts/DLS_install_prerequisites.sh
chmod +x DLS_install_prerequisites.sh
./DLS_install_prerequisites.sh

This script detects your Intel GPU/NPU, installs the correct drivers for Ubuntu 22.04 or 24.04, and adds your user to the required groups. Use --reinstall-npu-driver=yes to force-reinstall the NPU driver. Run ./DLS_install_prerequisites.sh --help for all options.

Step 2 — Install DL Streamer

Option A — Docker (recommended, zero setup):

# Run once on the host to allow X11 forwarding from containers
xhost +local:docker

docker run -it --rm \
  --device /dev/dri \
  --group-add $(stat -c "%g" /dev/dri/render*) \
  -e DISPLAY=$DISPLAY \
  -e XDG_RUNTIME_DIR=/tmp \
  -v /tmp/.X11-unix:/tmp/.X11-unix \
  intel/dlstreamer:latest

To use the NPU, also add --device /dev/accel --group-add $(stat -c "%g" /dev/accel/accel*) to the docker run command.

Option B — Native install (Ubuntu 24.04):

sudo -E wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB | gpg --dearmor | sudo tee /usr/share/keyrings/intel-gpg-archive-keyring.gpg > /dev/null
sudo -E wget -O- https://apt.repos.intel.com/edgeai/dlstreamer/GPG-PUB-KEY-INTEL-DLS.gpg | sudo tee /usr/share/keyrings/dls-archive-keyring.gpg > /dev/null
echo "deb [signed-by=/usr/share/keyrings/dls-archive-keyring.gpg] https://apt.repos.intel.com/edgeai/dlstreamer/ubuntu24 ubuntu24 main" | sudo tee /etc/apt/sources.list.d/intel-dlstreamer.list
sudo bash -c 'echo "deb [signed-by=/usr/share/keyrings/intel-gpg-archive-keyring.gpg] https://apt.repos.intel.com/openvino ubuntu24 main" | sudo tee /etc/apt/sources.list.d/intel-openvino.list'
sudo apt update && sudo apt-get install -y intel-dlstreamer

Full installation guide: Install Guide for Ubuntu | Windows

Quick Start - Run Your Pipeline

Step 1 — Set up models and environment (inside the container or on a native install):

cd ~
curl -LsSf https://astral.sh/uv/install.sh | sh
uv venv .dls-venv && source .dls-venv/bin/activate
uv pip install -r /opt/intel/dlstreamer/scripts/download_models/requirements_download_ultralytics_models.txt
python3 /opt/intel/dlstreamer/scripts/download_models/download_ultralytics_models.py \
  --model yolo11n.pt \
  --outdir ~/models/yolo11n \
  --int8
source /opt/intel/dlstreamer/scripts/setup_dls_env.sh

Step 2 — Run the pipeline. Change device=GPU to device=CPU or device=NPU — no other code changes needed.

gst-launch-1.0 \
  urisourcebin buffer-size=4096 uri=https://videos.pexels.com/video-files/1192116/1192116-sd_640_360_30fps.mp4 ! \
  decodebin3 ! \
  gvadetect model=~/models/yolo11n/yolo11n.xml device=GPU ! \
  queue ! \
  gvawatermark ! \
  gvafpscounter ! \
  videoconvert ! autovideosink sync=false

Output to JSON (works everywhere, including headless Docker):

gst-launch-1.0 \
  urisourcebin buffer-size=4096 uri=https://videos.pexels.com/video-files/1192116/1192116-sd_640_360_30fps.mp4 ! \
  decodebin3 ! \
  gvadetect model=~/models/yolo11n/yolo11n.xml device=GPU ! \
  queue ! \
  gvafpscounter ! \
  gvametaconvert format=json ! \
  gvametapublish file-format=json-lines file-path=output.json ! fakesink async=false

Python API

Create a file detect.py:

import gi
gi.require_version("Gst", "1.0")
from gi.repository import Gst
import os

Gst.init([])

video_url = "https://videos.pexels.com/video-files/1192116/1192116-sd_640_360_30fps.mp4"
model = os.path.expanduser("~/models/yolo11n/yolo11n.xml")
pipeline = Gst.parse_launch(f"""
    urisourcebin buffer-size=4096 uri={video_url} !
    decodebin3 !
    gvadetect model={model} device=GPU !
    queue !
    gvafpscounter !
    gvametaconvert format=json !
    gvametapublish file-format=json-lines file-path=output_from_python.json ! fakesink async=false
""")

pipeline.set_state(Gst.State.PLAYING)
bus = pipeline.get_bus()
bus.timed_pop_filtered(Gst.CLOCK_TIME_NONE, Gst.MessageType.EOS | Gst.MessageType.ERROR)
pipeline.set_state(Gst.State.NULL)

Then run it and wait for results:

python3 detect.py
# Detection results are written to output_from_python.json as the pipeline processes frames
# Each line is a JSON object with detected objects, labels, and bounding boxes
cat output_from_python.json

If any Python dependencies are missing, refer to the Python dependencies install guide.

DL Streamer Elements

Full elements reference →

Samples

30+ samples across Python, C++, and gst-launch command lines:

Browse all samples →

Supported Platforms

Operating systems: Ubuntu 22.04 / 24.04, Windows 11.

Full system requirements →

Documentation

Contributing

We welcome contributions! Please read CONTRIBUTING.md and follow the Code Style Guide.

For security issues, see SECURITY.md.

License

DL Streamer is licensed under the MIT License.

Intel, the Intel logo, OpenVINO, OpenVINO logo, Intel Geti, Intel Core, Intel Arc, and Intel Iris are trademarks of Intel Corporation or its subsidiaries. GStreamer is a trademark of the GStreamer project.