orchestrator-ready code for seamless integration
YOLO Detection • YOLO Pose • Tracking • ROI Analytics • Multi-Stream Pipelines • Python First
Fully Accelerated · Low Code · Docker Ready · Production Tested
| Component | Recommended / Supported |
|---|---|
| OS | Ubuntu 24.04 LTS |
| OpenVINO Version | 2025.4+ |
| DLStreamer Version | 2025.0.1.3 |
| Acceleration | Intel CPU • iGPU • Intel ARC |
| Docker Support | Yes – OpenVINO Runtime Containers |
| Bare Metal Support | Full |
✔️ Supports full XPU execution (CPU + GPU + VPU) ✔️ OAK-D / OAK-Lite / OAK-Pro support (DepthAI + OpenVINO backend) ✔️ Movidius MyriadX / NCS2 optimized ✔️ Multi-stream, multi-model, ROI pipelines ✔️ Python & C++
- 11th/12th/13th/14th Gen Intel Core
- Intel Xeon scalable
- Intel Atom (Edge)
- Intel UHD / Iris / Iris Xe
- Intel Xe-LP / Xe-HPG
- Tiger Lake, Alder Lake, Raptor Lake, Meteor Lake iGPU
- ARC A380 / A750 / A770
- ARC Pro Series
Install OpenVINO runtime → Clone repo → Run QuickTest.sh
Intel official quick install:
DLStreamer:
🔗 https://dlstreamer.github.io/get_started)
git clone https://github.com/bharath5673/DLstream.git
cd DLstream
bash QuickDemo.shRuns instantly with DLStreamer-ready pipelines:
- YOLO Detection
- YOLO Pose
- Multi-model, multi-stream pipelines
- Region-based analytics
- Full OpenVINO acceleration
- YOLO Models Converter
Run the entire AI video analytics stack inside Intel's official:
- OpenVINO Runtime Containers
- DLStreamer Media Analytics Containers
All pipelines run identical in Docker and Bare-Metal.
- Multi-model pipelines
- YOLO detection (OpenVINO IR / ONNX)
- YOLO-pose via OpenVINO
- Multi-stream tiled GStreamer pipelines
- Region-based analytics
- Python GStreamer bindings
- ❗C++ full application templates
- Triton-compatible (OpenVINO backend)
Minimal coding — edit configs and run.
You get:
- HIGH throughput using XPU execution
- ZERO CUDA / NVIDIA dependencies
- End-to-end pipelines optimized for CPU+iGPU
🔗 DLStreamer-Configs/MultiModel
🔗 DLStreamer-Python/ROI-counting
🔗 DLStreamer-Python/Pose/
🔗 DLStreamer-Python/tracking/
🔗 DLStreamer-Python/Pose/
🔗 DLStreamer-Python/face/
🔗 DLStreamer-Python/classification/
cd DLstream
bash QuickDemo.shDLstreamer/
├── demo
│ ├── pedestrian_tracker.gif
│ ├── segmentation.gif
├── demo.gif
├── DLStream-Configs
│ ├── dlstreamer_omz
│ │ ├── models
│ │ │ ├── intel
│ │ │ │ ├── age-gender-recognition-retail-0013
│ │ │ │ │ ├── FP16
│ │ │ │ │ │ ├── age-gender-recognition-retail-0013.bin
│ │ │ │ │ │ └── age-gender-recognition-retail-0013.xml
│ │ │ │ │ ├── FP16-INT8
│ │ │ │ │ │ ├── age-gender-recognition-retail-0013.bin
│ │ │ │ │ │ └── age-gender-recognition-retail-0013.xml
│ │ │ │ │ └── FP32
│ │ │ │ │ ├── age-gender-recognition-retail-0013.bin
│ │ │ │ │ └── age-gender-recognition-retail-0013.xml
│ │ │ │ ├── emotions-recognition-retail-0003
│ │ │ │ │ ├── FP16
│ │ │ │ │ │ ├── emotions-recognition-retail-0003.bin
│ │ │ │ │ │ └── emotions-recognition-retail-0003.xml
│ │ │ │ │ ├── FP16-INT8
│ │ │ │ │ │ ├── emotions-recognition-retail-0003.bin
│ │ │ │ │ │ └── emotions-recognition-retail-0003.xml
│ │ │ │ │ └── FP32
│ │ │ │ │ ├── emotions-recognition-retail-0003.bin
│ │ │ │ │ └── emotions-recognition-retail-0003.xml
│ │ │ ├── yolo11n_openvino_model
│ │ │ │ ├── n
│ │ │ │ │ ├── metadata.yaml
│ │ │ │ │ ├── yolo11n.bin
│ │ │ │ │ ├── yolo11n.json
│ │ │ │ │ └── yolo11n.xml
│ │ │ │ ├── yolo11s.bin
│ │ │ │ └── yolo11s.xml
│ │ │ ├── yolo11n.pt
│ │ │ ├── yolo11s
│ │ │ │ ├── FP16
│ │ │ │ │ ├── yolo11s.bin
│ │ │ │ │ └── yolo11s.xml
│ │ │ │ └── FP32
│ │ │ │ ├── yolo11s.bin
│ │ │ │ └── yolo11s.xml
│ │ ├── openvino_ga_cid
│ │ └── stats
│ └── download_YOLO_models.py
├── DLStream-Python
│ ├── dlstreamer_test_yolo.py
│ ├── dlstreamer_test_yolo_save_vid.py
│ ├── draw_face_attributes.py
│ └── hello_dlstreamer.py
├── inputs
│ ├── 1192116-sd_640_360_30fps.mp4
│ ├── head-pose-face-detection-female-and-male.mp4
│ └── youtube_stream_20250321_151616.mp4
├── Intel_logo.jpg
├── OpenVino.png
├── outputs
│ └── dlstream_output_20251209_072421.mp4
├── QuickDemo.sh ### main sh file
├── README.md
└── test_dl.py ### orchestartor running all python demos
Massive respect to the Intel Edge AI + OpenVINO community.
Pipelines, models, tracking logic, and deployment flows are inspired by best practices from DLStreamer and open-source AI/ML communities.
🟩 YOLO Ecosystem
🟦 Core AI / CV Architectures
- OpenVINO Model Zoo
- OpenCV DNN
- RepVGG, OREPA, FasterRCNN/SSD papers
- ONNX Runtime → OpenVINO conversion tools
🟧 Intel DLStreamer & Media Analytics
- DLStreamer (gst-gva)
- OpenVINO Execution Provider
- GStreamer plugins for inference (
gvadetect,gvaclassify,gvatrack, etc.)
Documentation: https://dlstreamer.github.io https://docs.openvino.ai
🔵 Tracking, ROI, Multi-Model Inspirations
- KLT + ByteTrack/DeepSORT concepts
- GVA ROI analytics
- Open-source MOT community
Thank you to every engineer and researcher contributing to
OpenVINO, DLStreamer, YOLO, tracking algorithms, and computer vision innovation.
This project stands on the shoulders of giants.