Opencv使用cuda实现图像处理

main.py

python 复制代码
import os
import cv2
print(f'OpenCV: {cv2.__version__} for python installed and working')
image = cv2.imread('bus.jpg')
if image is None:
    print("无法加载图像1")
print(cv2.cuda.getCudaEnabledDeviceCount())
cv2.cuda.setDevice(0)
cv2.cuda.printCudaDeviceInfo(0)
image_gpu = cv2.cuda_GpuMat()
image_gpu.upload(image)
screenshot = cv2.cuda.cvtColor(image_gpu, cv2.COLOR_RGB2BGR)
screenshot = cv2.cuda.resize(image_gpu, (400, 400))
result_cpu = screenshot.download()
print("图像宽度: ", image.shape)
print("数据类型:", image.dtype)
print("在CPU下,原始图像格式:", type(image))
print("在GPU下,处理后的图像格式:", type(screenshot))
print("在CPU下,处理后的图像格式:", type(result_cpu))
print("图像宽度: ", result_cpu.shape)
print("数据类型:", result_cpu.dtype)
if result_cpu is None:
    print("无法加载图像2")
else:
    cv2.imshow("Window", result_cpu)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

main.cpp

cpp 复制代码
#include <iostream>
#include <opencv2/opencv.hpp>
#include <opencv2/cudaimgproc.hpp>
int main(int argc, char* argv[])
{
	cv::Mat h_img1 = cv::imread("bus.jpg");
	cv::cuda::GpuMat d_result1, d_result2, d_result3, d_result4, d_img1;
	d_img1.upload(h_img1);
	cv::cuda::cvtColor(d_img1, d_result1, cv::COLOR_BGR2GRAY);
	cv::cuda::cvtColor(d_img1, d_result2, cv::COLOR_BGR2RGB);
	cv::cuda::cvtColor(d_img1, d_result3, cv::COLOR_BGR2HSV);
	cv::cuda::cvtColor(d_img1, d_result4, cv::COLOR_BGR2YCrCb);
	cv::Mat h_result1, h_result2, h_result3, h_result4;
	d_result1.download(h_result1);
	d_result2.download(h_result2);
	d_result3.download(h_result3);
	d_result4.download(h_result4);
	cv::imshow("Gray处理结果:", h_result1);
	cv::imshow("RGB处理结果:", h_result2);
	cv::imshow("HSV处理结果:", h_result3);
	cv::imshow("YCrCb处理结果:", h_result4);
	cv::waitKey(0);
	return 0;
}

CMakeLists.txt

bash 复制代码
cmake_minimum_required(VERSION 3.10)
project(OpenCVCuda LANGUAGES CXX)
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
set("OpenCV_DIR" "E:\\Opencv gpu\\newbuild\\install")
set(OpenCV_INCLUDE_DIRS ${OpenCV_DIR}\\include)
set(OpenCV_LIB_DEBUG ${OpenCV_DIR}\\x64\\vc17\\lib\\opencv_world470d.lib) 
set(OpenCV_LIB_RELEASE ${OpenCV_DIR}\\x64\\vc17\\lib\\opencv_world470.lib)   
set(CMAKE_CXX_STANDARD 14)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
set(CMAKE_CUDA_ARCHITECTURES 86)		
find_package(CUDA REQUIRED)			
enable_language(CUDA)  
include_directories(${OpenCV_INCLUDE_DIRS}) 
link_directories(${OpenCV_LIB_DIRS})  
find_package(OpenCV QUIET)	
link_libraries(${OpenCV_LIBS})
add_executable(OpenCVCuda main.cpp)
target_link_libraries(${PROJECT_NAME} 
    $<$<CONFIG:Debug>:${OpenCV_LIB_DEBUG}>
    $<$<CONFIG:Release>:${OpenCV_LIB_RELEASE}>
)
target_compile_features(OpenCVCuda PRIVATE cxx_std_14)
相关推荐
一切皆有可能!!5 小时前
实践篇:利用ragas在自己RAG上实现LLM评估②
人工智能·语言模型
月白风清江有声7 小时前
爆炸仿真的学习日志
人工智能
华奥系科技8 小时前
智慧水务发展迅猛:从物联网架构到AIoT系统的跨越式升级
人工智能·物联网·智慧城市
R²AIN SUITE8 小时前
MCP协议重构AI Agent生态:万能插槽如何终结工具孤岛?
人工智能
b***25119 小时前
动力电池点焊机:驱动电池焊接高效与可靠的核心力量|比斯特自动化
人工智能·科技·自动化
Gyoku Mint9 小时前
机器学习×第二卷:概念下篇——她不再只是模仿,而是开始决定怎么靠近你
人工智能·python·算法·机器学习·pandas·ai编程·matplotlib
小和尚同志9 小时前
通俗易懂的 MCP 概念入门
人工智能·aigc
dudly9 小时前
大语言模型评测体系全解析(下篇):工具链、学术前沿与实战策略
人工智能·语言模型
zzlyx999 小时前
AI大数据模型如何与thingsboard物联网结合
人工智能·物联网