5.22. sail.EngineImagePreProcess

Image inference interface with preprocessing function, internal use of thread pool way, higher efficiency while using Python.

5.22.1. __init__

Interface:

def __init__(self,
            bmodel_path: str,
            tpu_id: int
            use_mat_output: bool = False)

Parameters: * bmodel_path: str

The path of the input model.

• tpu_id: int

The tpu id that used.

• use_mat_output: bool

Whether to use OpenCV’s Mat as the output of the picture. The default value is False, which means not used.

5.22.2. InitImagePreProcess

Initialize the image preprocessing module.

Interface:

def InitImagePreProcess(self,
            resize_mode: sail_resize_type,
            bgr2rgb: bool = False,
            queue_in_size: int = 20,
            queue_out_size: int = 20) -> int

Parameters:

• resize_mode: sail_resize_type

Methods of internal scaling.

• bgr2rgb: bool

Whether to convert an image from BGR to GRB.

• queue_in_size: int

The maximum length of the input image queue cache, which defaults to 20.

• queue_out_size: int

The maximum length of Tensor queue cache of preprocess result , which is 20 by default.

Returns:

Return 0 on success and other values on failure.

5.22.3. SetPaddingAtrr

Sets the Padding properties, only works when the resize_mode is among BM_PADDING_VPP_NEAREST, BM_PADDING_TPU_NEAREST, BM_PADDING_TPU_LINEAR, BM_PADDING_TPU_BICUBIC.

Interface:

def SetPaddingAtrr(self,
            padding_b:int=114,
            padding_g:int=114,
            padding_r:int=114,
            align:int=0) -> int

Parameters: * padding_b: int

The padding pixel value of b channel, which defaults to 114.

• padding_g: int

The padding pixel value of g channel, which defaults to 114.

• padding_r: int

The padding pixel value of r channel, which defaults to 114.

• align: int

Image fill position, 0 indicates fill from the top left corner, 1 indicates center fill, default is 0.

Returns:

Return 0 on success and other values on failure.

5.22.4. SetConvertAtrr

Sets the properties of the linear transformation.

Interface:

def SetConvertAtrr(self, alpha_beta) -> int:

Parameters:

• alpha_beta: (a0, b0), (a1, b1), (a2, b2)。

  a0 is the coefficient of linear transformation for the 0th channel;

  b0 is the offset of linear transformation for the 0th channel;

  a1 is the coefficient of linear transformation for the 1th channel;

  b1 is the offset of linear transformation for the 1th channel;

  a2 is the coefficient of linear transformation for the 2th channel;

  b2 is the offset of linear transformation for the 2th channel;

Returns:

Return 0 on success and other values on failure.

5.22.5. PushImage

push image data

Interface:

def PushImage(self,
            channel_idx: int,
            image_idx: int,
            image: BMImage) -> int

Parameters: * channel_idx: int

The channel index of the input image

• image_idx: int

The image index of the input image

• image: BMImage

The input image

Returns:

Return 0 on success and other values on failure.

5.22.6. GetBatchData_Npy

Get a batch of inference results. When using this interface, use_mat_output must be False because the result type is BMImage.

Interface:

def GetBatchData_Npy(self)
-> tuple[[dict[str, ndarray], list[BMImage],list[int],list[int],list[list[int]]]]

Returns:

tuple[output_array, ost_images, channels, image_idxs, padding_attrs]

• output_array: dict[str, ndarray]

The inference result

• ost_images: list[BMImage]

Original image queue

• channels: list[int]

The result corresponds to the channel sequence of the original picture.

• image_idxs: list[int]

The result corresponds to the index sequence of the original picture.

• padding_attrs: list[list[int]]

The attribute list of the filling image. The starting point coordinate x, starting point coordinate y, the width after scaling, and the height after scaling.

5.22.7. GetBatchData_Npy2

Get a batch of inference results. When using this interface, use_mat_output must be True because the result type is numpy.ndarray[numpy.uint8].

Interface:

def GetBatchData_Npy2(self)
    -> tuple[dict[str, ndarray], list[numpy.ndarray[numpy.uint8]],list[int],list[int],list[list[int]]]

Returns:

tuple[output_array, ost_images, channels, image_idxs, padding_attrs]

• output_array: dict[str, ndarray]

The inference result

• ost_images: list[numpy.ndarray[numpy.uint8]]

Original image queue

• channels: list[int]

The result corresponds to the channel sequence of the original picture.

• image_idxs: list[int]

The result corresponds to the index sequence of the original picture.

• padding_attrs: list[list[int]]

The attribute list of the filling image. The starting point coordinate x, starting point coordinate y, the width after scaling, and the height after scaling.

5.22.8. GetBatchData

Get a batch of inference results. When using this interface, use_mat_output must be False because the result type is BMImage.

Interface:

def GetBatchData(self,
            need_d2s: bool = True)
            -> tuple[list[TensorPTRWithName], list[BMImage],list[int],list[int],list[list[int]]]

Parameters:

• need_d2s: bool

Whether to move data to the system memory. The default value is True, which means Yes.

Returns:

tuple[output_array, ost_images, channels, image_idxs, padding_attrs]

• output_array: list[TensorPTRWithName]

The inference result

• ost_images: list[BMImage]

Original image queue

• channels: list[int]

The result corresponds to the channel sequence of the original picture.

• image_idxs: list[int]

The result corresponds to the index sequence of the original picture.

• padding_attrs: list[list[int]]

The attribute list of the filling image. The starting point coordinate x, starting point coordinate y, the width after scaling, and the height after scaling.

5.22.9. get_graph_name

Get the name of model computation graph

Interface:

def get_graph_name(self) -> str

Returns:

Return the first name of model computation graph

5.22.10. get_input_width

Get the width of model input.

Interface:

def get_input_width(self) -> int

Returns:

Return the width of model input.

5.22.11. get_input_height

Get the height of model input.

Interface:

def get_input_height(self) -> int

Returns:

Return the height of model input.

5.22.12. get_output_names

Get tensor names of model output.

Interface:

def get_output_names(self) -> list[str]

Returns:

Return all tensor names of model output.

5.22.13. get_output_shape

Get the shape of the specified output Tensor

Interface:

def get_output_shape(self, tensor_name: str) -> list[int]

Parameters:

• tensor_name: str

The name of output tensor

Returns:

Return the shape of the specified output Tensor