opencv调用darknet框架图像分类

作者: wcdd | 来源:发表于2021-09-21 00:54 被阅读0次


///param img:输入图像
///param WeightsPath:权重文件路径
///param CfgPath:网络文件路径
int Classifier(Mat& img,string& WeightsPath,string& CfgPath)
{
    cv::dnn::ClassificationModel ClassNet(WeightsPath, CfgPath);///定义分类模型对象
    if (ClassNet.empty())
    {
        cout << "Can't load the net" << endl;
        return -1;
    }
    GPU计算
    //net.setPreferableBackend(dnn::DNN_BACKEND_CUDA);
    //net.setPreferableTarget(dnn::DNN_TARGET_CUDA_FP16);   
    ///CPU
    ClassNet.setPreferableBackend(dnn::DNN_BACKEND_OPENCV);
    ClassNet.setPreferableTarget(dnn::DNN_TARGET_CPU);
    ///输入图像预处理
    ClassNet.setInputSize(256, 256);///预处理，输入图像缩放
    ClassNet.setInputSwapRB(true);///RB通道数据交换
    ClassNet.setInputCrop(true);
    ClassNet.setInputScale(1.0 / 255.0);///输入数据归一化
 
 
    int id = 0;///类别号
    float c = 0.0;///识别概率
    ClassNet.classify(img, id, c);
}