Boosting Computer Vision: A Step-by-Step Guide to Compiling OpenCV with Python and CUDA Support on Rocky Linux 9

Are you ready to unlock the full potential of computer vision on your Rocky Linux 9 system? Look no further! In this comprehensive guide, we’ll walk you through the process of compiling OpenCV with Python and CUDA support, empowering you to harness the power of GPU acceleration for unparalleled performance.

What You’ll Need

Before we dive into the compilation process, make sure you have the following components installed and ready:

• Rocky Linux 9 (or a compatible Linux distribution)
• Python 3.9 or higher (we’ll use Python 3.9 as an example)
• NVIDIA GPU with CUDA capability (e.g., GeForce, Quadro, or Tesla)
• CUDA Toolkit (we’ll use version 11.5 as an example)
• cuDNN (optional but recommended for optimal performance)
• A decent internet connection for downloading required packages

Step 1: Install Required Dependencies

Let’s get started by installing the necessary dependencies for OpenCV compilation:

sudo dnf install -y epel-release
sudo dnf install -y cmake git make gcc-c++ pkgconfig
sudo dnf install -y libdc1394-devel libv4l-devel libxvidcore-devel
sudo dnf install -y libx264-devel ffmpeg-devel gstreamer1-plugins-base
sudo dnf install -y python39-devel numpy python39-pip

Installing CUDA Toolkit

Next, we’ll install the CUDA Toolkit, which is essential for GPU acceleration:

sudo dnf install -y https://developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9-11.5.0-1.x86_64.rpm
sudo dnf install -y cuda-toolkit-11-5

Installing cuDNN (Optional)

If you want to take advantage of NVIDIA’s cuDNN library for accelerated deep neural networks, follow these steps:

Download the cuDNN installation package from the NVIDIA website (cuDNN v8.2.4 for CUDA 11.5, in this case). Then, install it using the following commands:

sudo tar -xvf cudnn-11.5-linux-x64-v8.2.4.15.tgz
sudo cp cuda/include/cudnn.h /usr/local/cuda/include/
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64/
sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*

Step 2: Clone OpenCV Repository

Now, let’s clone the OpenCV repository from GitHub:

git clone https://github.com/opencv/opencv.git
git clone https://github.com/opencv/opencv_contrib.git

Step 3: Configure OpenCV Compilation

In the OpenCV repository, create a build directory and navigate into it:

mkdir opencv/build
cd opencv/build

Now, configure OpenCV compilation using CMake:

cmake -DCMAKE_BUILD_TYPE=RELEASE \
  -DCMAKE_INSTALL_PREFIX=/usr/local \
  -DINSTALL_C_EXAMPLES=ON \
  -DINSTALL_PYTHON_EXAMPLES=ON \
  -DOPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules \
  -DBUILD_EXAMPLES=ON \
  -DWITH_CUDA=ON \
  -DWITH_CUDNN=ON \
  -DCUDA_ARCH_PCIE=30 \
  -DCUDA_ARCH_BIN=6.0 \
  ../../

Here, we’re enabling:

• RELEASE build type for optimized performance
• Installation of C and Python examples
• Contribution modules from opencv_contrib
• Building examples
• CUDA support with cuDNN (if installed)
• Specify the CUDA architecture (Pascal, in this case)

Step 4: Compile and Install OpenCV

Now, let’s start the compilation process:

make -j$(nproc)
sudo make install

This might take some time, depending on your system’s processing power. Be patient!

Step 5: Verify OpenCV Installation

Once the installation is complete, let’s verify that OpenCV is working correctly:

python3.9 -c "import cv2; print(cv2.__version__)"

You should see the OpenCV version printed, indicating a successful installation.

Conclusion

Congratulations! You’ve successfully compiled OpenCV with Python and CUDA support on Rocky Linux 9. Your system is now ready to tackle computationally intensive computer vision tasks with the power of GPU acceleration.

Remember to update your system regularly to ensure you have the latest security patches and package updates. If you encounter any issues or have further questions, refer to the OpenCV documentation and community forums for assistance.

Bonus: Testing OpenCV with CUDA

Let’s test OpenCV with CUDA using a simple example:

import cv2
import numpy as np

# Create a sample image
img = np.random.rand(1024, 1024).astype(np.uint8)

# Convert to GPU matrix
gpu_img = cv2.cuda_GpuMat()
gpu_img.upload(img)

# Perform a simple operation (e.g., thresholding)
_, gpu_threshold = cv2.cuda.threshold(gpu_img, 127, 255, cv2.THRESH_BINARY)

# Download the result back to CPU
threshold = gpu_threshold.download()

# Display the result
cv2.imshow('Thresholded Image', threshold)
cv2.waitKey(0)
cv2.destroyAllWindows()

This code creates a random image, uploads it to the GPU, performs thresholding, and downloads the result back to the CPU for display.

Run this code and observe the significant performance boost provided by CUDA acceleration.

Component	Version
Rocky Linux	9
Python	3.9
CUDA Toolkit	11.5
cuDNN	8.2.4.15
OpenCV	4.5.5

This article provides a comprehensive guide to compiling OpenCV with Python and CUDA support on Rocky Linux 9. By following these steps, you’ll be able to harness the power of GPU acceleration for computer vision applications.

Remember to stay tuned for more tutorials and guides on computer vision and machine learning!

Frequently Asked Question

Get ready to unleash the power of OpenCV with Python and CUDA support on Rocky Linux 9!