问题描述

利用opencv-python的dnn模块调用yolo3模型进行目标检测的时候，根据网上的教程，加入GPU加速，也就是加入如下的两行代码：

 net.setPreferableTarget(cv::dnn::DNN_TARGET_CUDA);net.setPreferableBackend(cv::dnn::DNN_BACKEND_CUDA);

但是在运行之后，出现了这样的提示：

warning: setUpNet DNN module was not built with CUDA backend; switching to CPU

原因分析：

这说明GPU没被用上，原因是OPENCV在安装的时候，没有安装支持CUDA的版本，所以需要重新编译安装OPENCV。

解决方案：

安装OPENCV的教程可以参考这一个：https://blog.csdn.net/qq_37975595/article/details/109179388链接。注意，选择opencv4.4.0以上的版本！！我用的是4.4.0。本来我安装的是4.3.0，解决了一天的bug也没安装上！！！最后换了4.4.0，一下就装上了！

注意注意，cmake 这一步就是编译选项，我当时采用的是这个：

 cmake -DCMAKE_BUILD_TYPE=RELEASE \-DCMAKE_INSTALL_PREFIX=/usr/local \-DOPENCV_EXTRA_MODULES_PATH=../opencv_contrib-4.4.0/modules .. \-DWITH_CUDA=1 \-DCUDA_ARCH_BIN=7.5 \-DENABLE_FAST_MATH=1 \-DCUDA_FAST_MATH=1 \-DWITH_CUBLAS=1 \-DOPENCV_GENERATE_PKGCONFIG=1 \..

其中-DWITH_CUDA=1这句话就是选择了有CUDA版本的opencv。

-DCUDA_ARCH_BIN=7.5这个一定一定要选择跟你电脑显卡相匹配的数字，不然GPU还是不能用，还得重新编译安装。如果不匹配，会出现这样的报错：

 error: (-216:No CUDA support) OpenCV was not built to work with the selected device. Please check CUDA_ARCH_PTX or CUDA_ARCH_BIN in your build configuration. in function 'initCUDABackend'

此问题的解决参考了该博主的博客：https://blog.csdn.net/qq_34717531/article/details/108735629

warning: setUpNet DNN module was not built with CUDA backend； switching to CPU错误解决相关推荐

【Java】导入项目时，出现The project cannot be built until build path errors are resolved错误解决方法
[Java]导入项目时,出现The project cannot be built until build path errors are resolved错误解决方法参考文章: (1)[Java] ...
Android NDK: WARNING: Ignoring unknown import directory:错误解决方法
Android NDK: WARNING: Ignoring unknown import directory:错误解决方法添加NDK模块路径(参考,具体视情况): NDK_MODULE_PATH ...
AttributeError: module ‘tensorflow‘ has no attribute ‘Session‘错误解决
AttributeError: module 'tensorflow' has no attribute 'Session'错误解决根据网上教程安装tensorflow环境时,最后都会有个测试代码用 ...
Import error: no module named cv2 错误解决方法
Import error: no module named cv2 错误解决方法参考文章: (1)Import error: no module named cv2 错误解决方法 (2)https: ...
LNMP - Warning: require(): open_basedir restriction in effect错误解决方法
LNMP - Warning: require(): open_basedir restriction in effect错误解决方法参考文章: (1)LNMP - Warning: require ...
No module named scrapy 成功安装scrapy，却无法import的解决方法
No module named scrapy 成功安装scrapy,却无法import的解决方法参考文章: (1)No module named scrapy 成功安装scrapy,却无法impor ...
warning: control reaches end of non-void function：错误解决
warning: control reaches end of non-void function:错误解决用gcc编译一个程序的时候出现这样的警告: warning: control reache ...
ModuleNotFoundError: No module named ‘locust.events‘，httprunner执行locust压测报错解决办法~
ModuleNotFoundError: No module named 'locust.events',httprunner执行locust压测报错解决办法~ 问题背景上一篇文章讲解到初步使用ht ...
react-native error: Each module has to have a unique path.错误解决
1.Error:The modules'MOOC','android'point to same directory in the file system. Each module has to ha ...
pycharm Python3.6.5安装flask-No module named flask 错误解决
pycharm Python3.6.5安装flask No module named flask 错误解决 No module named flask 错误解决但发现Flask自己已经安装过Flas ...

warning: setUpNet DNN module was not built with CUDA backend； switching to CPU错误解决

问题描述

原因分析：

解决方案：

warning: setUpNet DNN module was not built with CUDA backend； switching to CPU错误解决相关推荐

最新文章

热门文章