Skip to content

README.md

Latest commit

 

History

History
277 lines (181 loc) · 7.92 KB

README.md

File metadata and controls

277 lines (181 loc) · 7.92 KB


DeepStream 8.0 – Ultra-Optimized AI Video Analytics Stack

🔖 EXCLUSIVE Release – Fully Optimized • Low-Code • Docker-Ready

YOLO Detection • YOLO Pose • Tracking • ROI Analytics • Multi-Stream Pipelines • Python First
Fully Optimized · Low Code · Docker Ready · Production Tested

🖥 Recommended System Setup

Component Recommended / Supported
OS Ubuntu 24.04 LTS
NVIDIA Driver 570.133.20
CUDA Compatibility Fully compatible with DeepStream 8.0
DeepStream Version DeepStream 8.0 (Production Ready)
Docker Support Yes – NVIDIA Container Runtime required
Bare Metal Support Supported (Native DS 8.0 Install)

✔️ Fully Docker Compatible ✔️ Supports Bare-Metal ✔️ Works for Python & C++ pipelines ✔️ Optimized for YOLOv5/YOLOv8/YOLO-Pose/Custom CNNs

⚡ Quick Start (1 Step)

Setup your GPU + environment → Pull repo → Run QuickTest.sh

Install NVIDIA driver

Follow NVIDIA official quick install:

🔗 https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_Quickstart.html

Clone this Repo and Run Quick Demo

git clone https://github.com/bharath5673/Deepstream.git

cd Deepstream
bash QuickDemo.sh

Runs instantly with DS8.0-ready configs:

  • YOLO Detection
  • YOLO Pose
  • Tracking
  • Multi-Model + Multi-Stream
  • ROI analytics

🎯 What This Repo Provides

✔️ Docker-Ready

Run your inference stack inside a fully isolated DeepStream 8.0 Docker environment. Just clone the prebuilt YOLO DS Docker image and start running demos instantly.

✔️ DeepStream 8.0 Templates (Production Ready)

  • Multi-model pipelines
  • YOLO detection & pose estimation
  • Trajectory tracking
  • ROI-based counting
  • Multi-stream tiled processing
  • Triton-ready configurations
  • Python & C++ implementations

✔️ Fully-Optimized & Low-Code

Minimal coding required — just edit config files and run. Get maximum performance with minimal effort.

🌟 Showcase Gallery

🔥 Multi-Model Pipeline

🔗 DeepStream-Configs/DeepStream-MultiModel

🟦 ROI Based Counting (Python)

🔗 DeepStream-Python/

🟧 Yolo POSE

🔗 DeepStream-Python/

⚙️ Custom CNN → DeepStream in 3 Steps

🔗 CNN-to-DeepStream/

⚡ Quick Demo

cd Deepstream
bash QuickDemo.sh

📂 Repo Structure

Deepstream/
│
├── DeepStream-Configs/
│   ├── DeepStream-MultiModel/
│   ├── test/ (multi-stream, tiling, custom pipelines)
│
├── DeepStream-Python/
│   ├── yolo
│   ├── yolo + pose
│   ├── ROI counting
│   ├── trajectory tracking
│
├── CNN-to-DeepStream/
│
└── QuickTest.sh

🙏 Acknowledgements

                               

Massive respect to the open-source community powering the DeepStream 8.0 ecosystem.
Models, configs, tracking logic, pose models, and deployment workflows are built on top of these amazing projects.

🔰 Credits & Sources

🟩 YOLO Ecosystem
🟦 Core AI / CV Architectures
🟧 NVIDIA + DeepStream + Metropolis
  • NVIDIA DeepStream SDK
  • NVIDIA Metropolis documentation
  • NVIDIA TensorRT & ONNX conversion tools
  • NVIDIA samples & reference apps
🔵 Tracking, ROI, Multi-Model Inspirations
  • NvDCF + KLT Tracker designs
  • MOT community publications
  • ROI analytics from DS sample apps
  • Common open-source tracking repos

⭐ Special Thanks

Thank you to every researcher, engineer, and developer who has contributed to
YOLO, tracking algorithms, CNN architectures, and DeepStream integration guides.

This project stands on the shoulders of giants.