# OpenCV — Open-Source Computer Vision Library

> The most widely used computer vision library with 2500+ optimized algorithms for image and video analysis, object detection, face recognition, and real-time processing across C++, Python, Java, and more.

## Install

Save as a script file and run:

# OpenCV — Open-Source Computer Vision Library

## Quick Use
```bash
pip install opencv-python
python -c "import cv2; img = cv2.imread('image.jpg'); print(img.shape)"
```

## Introduction
OpenCV (Open Source Computer Vision Library) is a BSD-licensed library with over 2500 optimized algorithms for computer vision and machine learning. Originally developed by Intel in 1999, it has become the standard toolkit for real-time image and video processing across academia and industry.

## What OpenCV Does
- Provides real-time image and video capture, processing, and analysis
- Offers object detection, face recognition, and tracking algorithms
- Supports camera calibration, 3D reconstruction, and stereo vision
- Includes deep learning inference via its DNN module for ONNX, TensorFlow, and PyTorch models
- Enables feature detection, optical flow, and image stitching

## Architecture Overview
OpenCV is written in C++ with bindings for Python, Java, and JavaScript (via OpenCV.js). The core modules handle matrix operations through the cv::Mat data structure. Hardware acceleration uses OpenCL for GPU compute and SIMD instructions for CPU optimization. The DNN module loads pretrained models from multiple frameworks without requiring their runtimes.

## Self-Hosting & Configuration
- Install via pip: `pip install opencv-python` or `opencv-contrib-python` for extra modules
- Build from source with CMake for custom module selection and GPU support
- Enable CUDA backend by setting `-D WITH_CUDA=ON` during CMake configuration
- Use `cv2.VideoCapture(0)` to access local cameras or RTSP streams
- Configure build flags for FFmpeg, GStreamer, or V4L2 video backend support

## Key Features
- Cross-platform: runs on Linux, macOS, Windows, iOS, and Android
- DNN module supports inference from ONNX, TensorFlow, Caffe, and Darknet models
- Real-time performance with optimized C++ core and hardware acceleration
- Extensive image processing: filters, morphology, color spaces, histograms
- Active community with 25+ years of development and broad ecosystem integration

## Comparison with Similar Tools
- **scikit-image** — Pure Python, better for prototyping but slower for real-time tasks
- **Pillow** — Simpler image I/O and manipulation but lacks computer vision algorithms
- **MediaPipe** — Focused on ML-based perception pipelines rather than classical CV
- **SimpleCV** — Higher-level API wrapping OpenCV but less maintained
- **Kornia** — Differentiable CV on PyTorch tensors for training pipelines

## FAQ
**Q: Does OpenCV require a GPU?**
A: No. OpenCV runs on CPU by default. GPU acceleration via CUDA or OpenCL is optional and requires building from source with the appropriate flags.

**Q: What is the difference between opencv-python and opencv-contrib-python?**
A: The contrib package includes extra modules like SIFT, SURF, ArUco markers, and additional deep learning layers not in the main package.

**Q: Can OpenCV run deep learning models?**
A: Yes. The DNN module loads ONNX, TensorFlow, Caffe, and Darknet models for inference without needing the original framework installed.

**Q: Is OpenCV suitable for production use?**
A: Yes. OpenCV is used in production at thousands of companies for real-time video analytics, robotics, autonomous vehicles, and medical imaging.

## Sources
- https://github.com/opencv/opencv
- https://docs.opencv.org/

---
Source: https://tokrepo.com/en/workflows/96185804-3c91-11f1-9bc6-00163e2b0d79
Author: Script Depot