Cropping

Function Description

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

For details about the API, see Crop.

API Calling Process

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

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

Key steps are demonstrated as follows:

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

    Select the one-to-one or one-to-many image cropping mode based on service requirements to construct the corresponding input Rect and output tensor.

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

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

    • Asynchronous execution
      1. Create a stream. For details, see Asynchronous Invocation.
      2. Transfer the input image, created stream, and other parameters to the Crop method to obtain the image cropping 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
 
// 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.
    MxBase::Rect rect(0, 0, 320, 320);

    // Construct an output tensor.
    std::vector<uint32_t> dstShape = {320, 320, 3};
    MxBase::Tensor dst(dstShape, MxBase::TensorDType::UINT8, -1);
    dst.Malloc();
    dst.ToDevice(0);

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

    // Perform cropping.
    APP_ERROR ret = MxBase::Crop(input, rect, dst, keepMargin);
    if (ret != APP_ERR_OK) {
        std::cout << "Crop failed." << std::endl;
    }
}
// Deinitialization
MxBase::MxDeInit();
Parent topic: Image Processing Using Tensors (Tensor)