OpenCV 和 TensorRT 之间的数据转换 HWC to CHW-526互联

TensorRT做图像相关模型部署的时候，导入图片的数据存储往往是BHWC(Batch, Height, Width, Channel), 而TensorRT推理的时候是BCHW. OpenCV 和 TensorRT 之间的数据转换(BHWC to BCHW)，一般是所有元素遍历赋值：

    cv::Mat origin_image = cv::imread("test.jpg", 1);
    if (!origin_image.data)
    {
        cerr << "Error : could not load image." << endl;
        return false;
    }
    cv::Mat resize_image;
    cv::resize(origin_image, resize_image, cv::Size(inputH, inputW), cv::INTER_CUBIC);

    // Fill data buffer
    float* hostDataBuffer = static_cast<float*>(buffers.getHostBuffer(mParams.inputTensorNames[0]));
    for (int i = 0, volImg = inputC * inputH * inputW; i < batchSize; ++i)
    {
        for (int c = 0; c < inputC; ++c)
        {
            for (unsigned j = 0, volChl = inputH * inputW; j < volChl; ++j)
                hostDataBuffer[i * volImg + c * volChl + j] = 1.0 - float(resize_image.data[j * inputC + 2 - c]/255.);
        }
    }

这种方式简单易懂，不过还可以用OpenCV的方式来做数据转换，效率也会提高很多。例如再GPU上操作，用OpenCV里面的split把图片分3通道，然后直接整块赋值：

    // example 2
    float* deviceDataBuffer = static_cast<float*>(buffers.getDeviceBuffer(mParams.inputTensorNames[0]));
    std::vector<cv::cuda::GpuMat> chw;
    for(size_t i = 0 , volImg = inputC*inputH*inputW, volChl = inputH*inputW; i < batchSize; i++ ) {
        std::vector<cv::cuda::GpuMat> chw {
            cv::cuda::GpuMat(inputH, inputW, CV_32F, &deviceDataBuffer[i*volImg + 0*volChl]),
            cv::cuda::GpuMat(inputH, inputW, CV_32F, &deviceDataBuffer[i*volImg + 1*volChl]),
            cv::cuda::GpuMat(inputH, inputW, CV_32F, &deviceDataBuffer[i*volImg + 2*volChl])
        };
        cv::cuda::split(G_images[i], chw); // G_images是多张图片，数据类型是CV::cuda::GpuMat
    }