Resizing

Function Description

Resizes the input image and outputs it to the tensor object.

For details about the API, see Resize.

API Calling Process

Before calling the resizing APIs, prepare the image to be resized and convert it to a tensor object.

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

Key steps are demonstrated as follows:

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

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

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

     No stream is created. The input image and other parameters are passed to the Resize method to obtain the image 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 Resize method to obtain the image resizing result.
4. Deinitialize the initialized global resources by calling MxDeInit().

Sample Code

The following is a code example of key steps of functions and features, which is for reference only and cannot be directly copied for compilation or running.

 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

// 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 resizing width and height.
    MxBase::Size resize(289, 289);

    // Construct an output tensor.
    std::vector<uint32_t> dstShape = {289, 304, 3};
    MxBase::Tensor dst(dstShape, MxBase::TensorDType::UINT8, -1);
    dst.Malloc();
    dst.ToDvpp(0);

    // Whether to reserve invalid boundaries of the input tensor for the output tensor
    bool keepMargin = true;

    // Perform resizing.
    APP_ERROR ret = MxBase::Resize(input, dst, resize, MxBase::Interpolation::BILINEAR_SIMILAR_OPENCV, keepMargin);
    if (ret != APP_ERR_OK) {
        std::cout << "Resize failed." << std::endl;
    }
}

// Deinitialization
MxBase::MxDeInit();