ccm-edge-cv-pipeline

A modular, lightweight, C++ computer-vision pipeline designed for Jetson, Raspberry Pi, and general Linux systems.
Built for real-time processing, config-driven pipelines, and extensibility.

Work-in-progress: This repository is part of the CCMCode Embedded Vision toolkit and serves as a portfolio-quality example of modern C++ pipeline architecture.

🚀 Status Overview

This repository is under active development.
It already includes a minimal working pipeline, and the architecture is evolving toward a fully configurable multi-stage system.

✔ Implemented Features (Current)

🎥 Basic Camera Pipeline

• OpenCV capture from USB/MIPI (index/size from config)
• Grayscale + simple processing stages
• Real-time FPS counter
• Clean separation between app, core, I/O, and stages

🧩 Minimal Modular Architecture

• IPipelineStage interface
• Simple Pipeline/config loading core
• First concrete stage: GrayscaleStage
• IO abstraction layer (ICameraSource) with an OpenCV-based implementation

⚙️ Initial YAML Configuration Support

• Configure:
  • camera index
  • resolution
  • FPS target
  • enabled stages
• YAML → pipeline construction (basic version)

🛠 Modern CMake Build

• CMake 3.16+
• C++17
• Targets:
  • edgecv_core
  • edgecv_app
• OpenCV detection + linking

📍 Planned Features (Roadmap)

🔁 Pipeline & Architecture

• Async / threaded stages
• Back-pressure system
• Frame pool allocator
• Stage-level metrics + profiling hooks

🧠 AI & Acceleration

• ONNX models
• TensorRT integration (Jetson)
• SIMD acceleration for common operators

🏗 Deployment

• Jetson-optimized Docker image
• Raspberry Pi tuned build flags

📦 I/O Backends

• GStreamer capture nodes
• RTSP streaming output
• Video recorder stage

📂 Repository Layout (Current + Planned)

ccm-edge-cv-pipeline/
├── CMakeLists.txt
├── config/
│   └── pipeline_basic.yaml
├── docs/
│   └── architecture.md
├── include/
│   ├── core/
│   │   ├── config_loader.hpp
│   │   ├── ipipeline_stage.hpp
│   │   └── perf_timer.hpp
│   ├── io/
│   │   ├── icamera_source.hpp
│   │   └── opencv_camera.hpp
│   └── stages/
│       └── grayscale_stage.hpp
├── src/
│   ├── app/
│   │   └── main.cpp
│   ├── core/
│   │   └── config_loader.cpp
│   ├── io/
│   │   ├── camera_capture.hpp
│   │   └── opencv_camera.cpp
│   └── stages/
│       └── grayscale_stage.cpp
└── tests/
    └── test_frame_queue.cpp

ℹ️ Some directories represent the target architecture and are partially implemented.

🧪 Example YAML Configuration

config/pipeline_basic.yaml:

camera:
  index: 0
  width: 640
  height: 480
  fps: 30

pipeline:
  stages:
    - type: grayscale

▶ Building and Running

1. Configure

mkdir build && cd build
cmake -DEDGECV_ENABLE_TENSORRT=OFF -DEDGECV_ENABLE_PROFILING=ON ..

2. Build

cmake --build .

3. Run

./edgecv_app ../config/pipeline_basic.yaml

🏗 Architecture Overview

See full details in docs/architecture.md.

[ Camera Capture (ICameraSource / OpenCV camera) ]
        ↓
[ Frame Queue / Capture Loop ]
        ↓
[ Pipeline (Stage Chain) ]
        ↓
[ Output (Window / File / Stream) ]

🧪 Tests

test_frame_queue.cpp

• Push/pop behaviour
• Blocking wait
• Thread-safety basics

Run tests:

ctest

🔧 CI / CD

Badges:

🔗 Related Projects

CCM ESP32 Vision Node

MCU camera streaming node for edge devices.

CCM EdgeVision Starter Kit

Jetson/Pi YOLO & GStreamer CV starter repo.

👨‍💻 About CCMCode

This repository forms part of CCMCode’s Edge Vision Suite, a collection of modern, minimal, and production-oriented embedded CV systems.

📜 License

MIT License.

---

🚀 Professional Edge Pipelines

This is the open-source foundation of the CCMCode Vision Stack. It focuses on modular, non-CUDA functionality for broad compatibility.

Looking for the Pro SDK? I am developing a high-performance Middleware SDK for advanced non-CUDA optimization and enterprise deployment.

• Sponsor to support open features: GitHub Sponsors
• Inquire about the Private SDK: info@ccmcode.dev