Cropping and Resizing

Function Description

Crops and resizes the input image, and outputs the image to the tensor object.

For details about the API, see CropResize.

API Calling Process

Before calling the cropping and resizing API, prepare an image to be cropped and resized and convert it to a tensor object.

Figure 1 Process of calling APIs in tensor mode (cropping and resizing)

Key steps are demonstrated as follows:

1. Perform global initialization by calling MxInit().
2. Construct the cropping and resizing parameters and the output tensor.

   Construct the input Rect and Size as well as the output Tensor based on service requirements.

3. Select the synchronous or asynchronous cropping and resizing mode based on the service requirements.
   • Synchronous execution

     No stream is created. The input image and other parameters are passed to the CropResize method to obtain the image cropping and resizing result.

   • Asynchronous execution
     1. Create a stream. For details, see Asynchronous Invocation.
     2. Transfer the input image, created stream, and other parameters to the CropResize method to obtain the image cropping and resizing result.
4. Deinitialize the initialized global resources by calling MxDeInit().

Sample Code

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35

// Initialization
MxBase::MxInit();

{
    // Image reading
    std::string imgPath = "./test.jpg";
    cv::Mat imgData = cv::imread(imgPath, cv::IMREAD_UNCHANGED);
    std::vector<uint32_t> shape{600, 600, 3};

    // Construct an input tensor.
    MxBase::Tensor input(imgData.data, shape, MxBase::TensorDType::UINT8, -1);
    input.ToDvpp(0);

    // Set the cropped area and resizing width and height.
    MxBase::Rect rect(0, 0, 320, 320);
    std::vector<MxBase::Rect> cropRectVec = {rect};
    MxBase::Size size(480, 480);
    std::vector<MxBase::Size>sizeVec = {size};

    // Construct an output tensor.
    std::vector<uint32_t> dstShape = {480, 480, 3};
    MxBase::Tensor dst(dstShape, MxBase::TensorDType::UINT8, -1);
    dst.Malloc();
    dst.ToDvpp(0);
    std::vector<MxBase::Tensor> outputTensorVec = {dst};

    // Perform cropping and resizing.
    APP_ERROR ret = MxBase::CropResize(input, cropRectVec, sizeVec, outputTensorVec, MxBase::Interpolation::BILINEAR_SIMILAR_OPENCV, true);
    if (ret != APP_ERR_OK) {
        std::cout << "CropResize failed." << std::endl;
    }
}

// Deinitialization
MxBase::MxDeInit();