CN112037157A - Data processing method and device, computer readable medium and electronic equipment - Google Patents

Abstract

The present disclosure relates to the technical field of data processing, and in particular, to a data processing method, a data processing apparatus, a computer-readable medium, and an electronic device. The method includes: acquiring a feature map to be processed, and inputting the feature map into an expansion rate calculation layer to obtain an expansion rate table corresponding to the feature map; wherein the expansion rate table includes each pixel in the feature map The corresponding expansion rate value; input the expansion rate table into the expansion convolution layer, so that the expansion convolution layer combines the expansion rate table to perform expansion convolution processing on the feature map, so as to obtain the corresponding expansion rate of the feature map. characteristic data. The method can realize the adaptive and fine adjustment of the receptive field of each pixel in the feature map, avoid the omission of features in the feature extraction process, and can improve the sampling density and retain the detailed features.

Description

Data processing method and apparatus, computer readable medium and electronic device

technical field

The present disclosure relates to the technical field of data processing, and in particular, to a data processing method, a data processing apparatus, a computer-readable medium, and an electronic device.

Background technique

When using deep learning algorithms for image processing, expanding the receptive field helps the neural network to receive a wider range of information and extract more complex information. In general, a conventional way to expand the receptive field is to increase the number of layers of the neural network, and the deeper the network layer, the wider the information is received. However, its disadvantage is that the network complexity is greatly increased, and the difficulty of convergence is also increased. Another way to dilate convolution is to expand the receptive field by expanding the filter size. However, the disadvantage is that the convolution operation regularly skips some pixels, which will form holes in the generated features, resulting in features being missed.

It should be noted that the information disclosed in the above Background section is only for enhancement of understanding of the background of the present disclosure, and therefore may contain information that does not form the prior art that is already known to a person of ordinary skill in the art.

SUMMARY OF THE INVENTION

The present disclosure provides a data processing method, a data processing apparatus, a computer-readable medium, and a terminal device, which can adaptively adjust the receptive field during the calculation process of the dilated convolution to avoid feature omission.

Other features and advantages of the present disclosure will become apparent from the following detailed description, or be learned in part by practice of the present disclosure.

According to a first aspect of the present disclosure, there is provided a data processing method, comprising:

Obtain the feature map to be processed, and input the feature map into the expansion rate calculation layer to obtain the expansion rate table corresponding to the feature map; wherein, the expansion rate table includes the expansion rate value corresponding to each pixel in the feature map ;

The dilation rate table is input into the dilation convolution layer, so that the dilation convolution layer performs dilation convolution processing on the feature map in combination with the dilation rate table, so as to obtain feature data corresponding to the feature map.

According to a second aspect of the present disclosure, there is provided a data processing apparatus, comprising:

The adaptive expansion rate table calculation module is used to obtain the feature map to be processed, and input the feature map into the expansion rate calculation layer to obtain the expansion rate table corresponding to the feature map; wherein, the expansion rate table includes the The expansion rate value corresponding to each pixel in the feature map;

a dilated convolution processing module, configured to input the dilation rate table into the dilated convolution layer, so that the dilated convolution layer combines the dilation rate table to perform dilated convolution processing on the feature map to obtain the feature The feature data corresponding to the graph.

According to a third aspect of the present disclosure, there is provided a computer-readable medium on which a computer program is stored, and when the computer program is executed by a processor, implements the above-mentioned data processing method.

According to a fourth aspect of the present disclosure, there is provided a terminal device, comprising:

one or more processors;

A storage device for storing one or more programs, which when executed by the one or more processors, enables the one or more processors to implement the above-mentioned data processing method.

In the data processing method provided by an embodiment of the present disclosure, the expansion rate calculation layer is configured, and the expansion rate convolution layer is used to calculate the expansion rate table corresponding to the feature map; Query the expansion rate table to obtain the expansion rate value corresponding to each pixel, and perform sampling according to the corresponding expansion rate value of each pixel, so as to realize the adaptive and refined adjustment of the receptive field of each pixel in the feature map, and avoid the process of feature extraction. The omission of features; and can improve the sampling density and retain detailed features.

It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description serve to explain the principles of the disclosure. Obviously, the drawings in the following description are only some embodiments of the present disclosure, and for those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

1-1 schematically shows a schematic diagram of a dilated convolution sampling with a dilation rate of 1 in an exemplary embodiment of the present disclosure;

1-2 schematically show a schematic diagram of a dilated convolution sampling with a dilation rate of 2 in an exemplary embodiment of the present disclosure;

1-3 schematically show a schematic diagram of a dilated convolution sampling with a dilation rate of 4 in an exemplary embodiment of the present disclosure;

FIG. 2 schematically shows a schematic flowchart of a data processing method in an exemplary embodiment of the present disclosure;

FIG. 3 schematically shows a schematic flowchart of a method for dilating a convolutional layer in an exemplary embodiment of the present disclosure;

FIG. 4 schematically shows a schematic diagram of a sampling result of a dilated convolution layer in an exemplary embodiment of the present disclosure;

FIG. 5 schematically shows a schematic composition diagram of a display control apparatus in an exemplary embodiment of the present disclosure;

FIG. 6 schematically shows a schematic structural diagram of an electronic device of a terminal device in an exemplary embodiment of the present disclosure.

Detailed ways

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments, however, can be embodied in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus their repeated descriptions will be omitted. Some of the block diagrams shown in the figures are functional entities that do not necessarily necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor devices and/or microcontroller devices.

When using deep learning algorithms to process images, such as in image noise reduction, image recognition and other application scenarios, when using convolutional neural networks to extract features from images, expanding the receptive field helps the neural network to receive a wider range of information and extract more complex information. A conventional way to expand the receptive field is to increase the number of layers of the neural network, the deeper the network layer receives, the wider the information. The disadvantage is that the network complexity is greatly increased, and the difficulty of convergence is also increased. The principle of dilated convolution is to inject 0 into the convolution kernel to expand the sampling range of the convolution operation without changing the computational complexity. For example, refer to the schematic diagrams of dilated convolution sampling shown in Figure 1-1 to Figure 1-3, where the dilation rate of Figure 1-1 is 1, the convolution kernel is the original form, and the sampling points are closely connected. Figure 1-2 has a dilation rate of 2, the convolution kernel spacing is injected with zeros, and the valid samples are spaced apart. Figure 1-3 has an expansion rate of 4 and a spacing of 4 between valid sampling points. The advantage of dilated convolution is to expand the receptive field, but it will miss subtle features. Another common way is to use dilated convolution. The concept of dilated convolution is to expand the size of the filter and fill it with zeros at a certain interval. The expansion ratio of the size is the expansion rate. The disadvantage is that some pixels are regularly skipped during the convolution operation, which is easy to form holes in the generated features.

In view of the above-mentioned shortcomings and deficiencies of the prior art, the present exemplary embodiment provides a data processing method. Referring to Fig. 2, the above-mentioned data processing method may include the following steps:

S11: Acquire the feature map to be processed, and input the feature map into the expansion rate calculation layer to obtain an expansion rate table corresponding to the feature map; wherein the expansion rate table includes the expansion corresponding to each pixel in the feature map rate value;

S12: Input the dilation rate table into the dilation convolution layer, so that the dilation convolution layer performs dilation convolution processing on the feature map in combination with the dilation rate table, so as to obtain feature data corresponding to the feature map.

In the data processing method provided by this example embodiment, the expansion rate calculation layer is configured, and the expansion rate convolution layer is used to calculate the expansion rate table corresponding to the feature map; so that when the expansion convolution calculation is performed on the feature map, the expansion rate can be queried. The expansion rate value corresponding to each pixel is obtained from the rate table, and sampling is performed according to the corresponding expansion rate value of each pixel. The omission of features; on the other hand, it can improve the sampling density and retain detailed features.

Hereinafter, each step of the data processing method in this exemplary embodiment will be described in more detail with reference to the accompanying drawings and embodiments.

In step S11, the feature map to be processed is obtained, and the feature map is input into the expansion rate calculation layer to obtain the expansion rate table corresponding to the feature map; wherein, the expansion rate table includes each pixel in the feature map The corresponding expansion rate value.

In this exemplary embodiment, the above-mentioned feature map to be processed may be a feature map obtained after one feature extraction is performed on the original image. For example, in image noise reduction, image recognition, and image registration, the feature map obtained after feature extraction is performed on the input original image.

The above data processing method can provide a dilation rate calculation layer and a dilation convolution layer. Specifically, after acquiring the feature map to be processed, it can be input into the dilation rate convolution layer to calculate the dilation rate value corresponding to each pixel in the feature map, and obtain the dilation rate table corresponding to the feature map.

Specifically, the expansion rate calculation layer can be one or a plurality of convolutional layers set in succession, and the feature map is subjected to one or more consecutive convolution processing to obtain the expansion rate table. Realize the use of the pixel features of each pixel in the feature map to adaptively set the expansion rate for each pixel. Alternatively, in other exemplary embodiments of the present disclosure, when the expansion rate calculation layer includes a plurality of convolutional layers, a pooling layer (pooling) and a residual network module may also be included; The basic components of neural networks such as image scaling modules for dimensionality reduction processing.

Also, the image size of the dilation rate table can be configured according to the resolution of the feature map or the image size. Alternatively, the image size of the dilation rate table can also be configured by combining the information of the feature maps and the dilated convolutional layers. For example, the image size of the dilation rate table can be configured according to the image size of the feature map; or, the dilation rate table can be configured by combining the image size of the feature map and the number of output channels of the dilated convolutional layer or configure the image size of the dilation rate table by combining the image size of the feature map and the filter grouping information of the output channel of the dilated convolution layer.

Specifically, the image size (or resolution) of the expansion rate table can be set to be the same as the image size (or resolution) of the feature map; for example, the image size of the input feature map is width (width)*height (height)*input The number of channels (channel_in), the image size of the expansion rate table is also Width*height*channel_in. Alternatively, set the image size of the dilation rate table to be the product of the image size of the feature map and the number of output channels of the dilated convolution layer. For example, the input feature map image size is Width*height*channel_in, and the output channel of the dilated convolutional layer The number is channel_out, that is, the output feature map image size is Width*height*channel_out, and the image size of the expansion rate array is Width*height*(channel_in*channel_out). Or, set the image size of the dilation rate table to be the product of the image size of the feature map and the number of filter groups in the output channel number of the dilated convolution layer; for example, the input feature map image size is Width*height*channel_in, for The filter bank corresponding to each output channel of the dilated convolution layer is divided into several groups, and the number of groups is group, and the image size of the dilation rate is Width*height*channel_in*group.

In step S12, the dilation rate table is input into the dilation convolution layer, so that the dilation convolution layer combines the dilation rate table to perform dilation convolution processing on the feature map, so as to obtain the corresponding data of the feature map. characteristic data.

In this example implementation, referring to FIG. 3 , the above-mentioned step S12 may include:

Step S121, query the expansion rate table to obtain the expansion rate value corresponding to each pixel in the feature map.

Specifically, after calculating the dilation rate table corresponding to the feature map, both the feature map and the dilation rate table can be input into the dilated convolution layer. The expansion rate table is queried by the expansion convolution layer to obtain the expansion rate value corresponding to each pixel, and the sampling range of the convolution operation is adjusted according to the expansion rate value, and then the convolution operation is performed to process the features. Wherein, based on the corresponding dilation rate value of each pixel, sampling is performed using the same interval in the horizontal direction and the vertical direction of the pixel. That is, when the dilation rate value corresponding to one pixel is d, the interval between the sampling points of the dilated convolution operation in the horizontal direction and the vertical direction is both d.

In some exemplary embodiments of the present disclosure, the expansion rate value d in the expansion rate table may be an integer, or a non-integer (decimal). When the expansion rate value is an integer, the corresponding expansion rate value can be used directly.

If the expansion rate value d corresponding to the pixel in the feature map is a decimal, the coordinates of the sampling point may be decimals. Therefore, an interpolation operation unit can also be set in the dilated convolution layer, which is used to estimate the feature value of the sampling point by using the pixel interpolation of the neighborhood when the dilation rate value is non-integer, and then perform the convolution operation of the dilated convolution layer. . Specifically, it can include:

Step S21, obtaining the neighborhood pixel value of the sampling point corresponding to the pixel, and calculating the corresponding neighborhood pixel interpolation value based on the neighborhood pixel value;

Step S22, estimating the eigenvalues of the sampling points according to the neighborhood pixel interpolation, so as to perform dilation convolution based on the eigenvalues of the sampling points.

For example, the bi-line interpolation method can be used to calculate the weight of each pixel according to the distance by using this number of coordinate pixels in the specified neighborhood, and estimate the eigenvalues at the sampling coordinates. Specifically, the formula can include:

Among them, x is the received feature map, y is the output feature map, P ₀ is the coordinate position on the output feature map y, w represents the convolution kernel of the dilated convolution layer when the dilation rate is 1, and R represents the w All coordinate positions of , P _n enumerates the coordinate positions in R, G() is the interpolation weight function, d represents the calculated dilation rate, (P ₀ +d · P _n ) enumerates the dilation convolution layer at the dilation rate The coordinate position of the sampling point at d, N represents the integer coordinate position of the neighborhood of the sampling point, and q enumerates the integer coordinates in N.

G(q, P ₀ +d·P _n ) in the above formula calculates the interpolation weight according to the spatial distance between q and (P ₀ +d·P _n ), and the interpolation weight is inversely proportional to the spatial distance. Optionally, G( ) adopts a bilinear interpolation method or a bicubic interpolation method. N in the above formula represents the integer coordinate position in the neighborhood of the sampling point. Optionally, N is the integer coordinate position in the 4*4 neighborhood of the sampling point. That is, if the coordinates of the sampling point are (row,col), and row and col can be approximated as integers row', col', then N represents from the (row-1)th row to the (row+2)th row, from the (row-1)th row to the (row+2)th row. Four rows and four columns between column col-1) and column (col+2), 16 integer coordinate positions. Referring to the embodiment shown in Figure 4, the expansion rate d may be a decimal, resulting in the sampling point coordinates may be decimals. Therefore, the pixel interpolation of the neighborhood is used to estimate the eigenvalues of the sampling points, and then convolve with the expansion convolution layer. operation.

As a reference comparison, the conventional convolutional layer calculation formula is usually:

Here w refers to the weight value of the convolution kernel, P ₀ refers to the center pixel coordinate to be calculated, and P _n enumerates the pixel coordinate position in the neighborhood. When the size of the convolution kernel is 3x3, _Pn represents ₉ pixels in the 3*3 neighborhood of P0. The conventional dilated convolution formula is usually:

Compared with the above conventional convolution, the expansion coefficient d is increased. The physical meaning is that the spacing of sampling points changes from 1 to d.

Compared with the above-mentioned prior art, the embodiment of the present disclosure adds an interpolation function G( ) and an integer coordinate q. The physical meaning is that the pixel value of the sampling point needs to be estimated by interpolation according to the pixel value x(q) at the integer coordinate q of the neighborhood. It can be ensured that when the expansion rate value in the expansion rate table is a decimal, the corresponding expansion rate value can be accurately obtained for sampling.

Referring to FIG. 3 , the above-mentioned step S12 may further include:

Step S122: Determine the sampling range of the corresponding pixel according to the dilation rate value, so as to perform dilation convolution processing.

In some exemplary embodiments of the present disclosure, when the dilation rate table is the same as the image size of the input feature map, for each layer output channel of the dilated convolutional layer, the same coordinate pixel is used in each layer output channel. The dilation rate value is dilated and convoluted.

For example, when the image size of the feature map and the expansion rate table are the same, (Width*height*channel_in), and channel_in=2, the expansion rate table contains two layers of expansion rate sub-tables, corresponding to the two layers of the feature map respectively. A channel of one dimension, and the two expansion rate subtables are the same. When the number of output channels of the dilated convolution calculation layer is channel_out=3, at this time, the pixels of the first layer channel of the feature map query the dilation rate table to obtain the corresponding dilation rate value, determine the interval according to the dilation rate value and perform sampling, and then proceed to Convolution calculation to obtain the first feature value; and, query the expansion rate table for the pixels of the second layer channel of the feature map to obtain the corresponding expansion rate value, determine the interval according to the expansion rate value and perform sampling, and then perform convolution calculation to obtain The second eigenvalue; the eigenvalue of each pixel of the output channel of the first layer of the dilated convolution layer is determined according to the first eigenvalue and the second eigenvalue. By analogy, a feature map with 3 output channels of the dilated convolutional layer can be obtained.

In some exemplary embodiments of the present disclosure, different dilation rates may also be calculated for filter banks corresponding to each output channel of the dilated convolution layer. That is, the image size of the dilation rate table is the product of the image size of the feature map and the number of output channels of the dilated convolution layer. For example, the size of the input feature map image is Width*height*channel_in, and the number of output channels of the dilated convolutional layer is channel_out, that is, the size of the output feature map image is Width*height*channel_out, and the image of the dilation rate table at this time The dimensions are Width*height*channel_in*channel_out. For example, the input feature map image size is Width*height*channel_in, channel_in=2; and the number of output channels of the dilated convolutional layer is channel_out=3, the image size of the dilation rate table is Width*height*6. At this time, the channel pixels of the first layer dimension of the feature map query the expansion rate value of the first layer dimension of the table for sampling, and the first feature value is obtained after the convolution calculation; the channel pixels of the second layer dimension query the expansion rate The dilation rate value of the second layer dimension of the table is sampled, and the second eigenvalue is obtained after the convolution calculation; the eigenvalue of the pixel of the output channel of the first layer of the dilated convolution layer is obtained according to the first eigenvalue and the second eigenvalue. . The channel pixels of the first layer dimension of the feature map query the expansion rate table of the third layer dimension to sample the expansion rate value of the third layer dimension, and obtain the third feature value after convolution calculation; the channel pixels of the second layer dimension query the expansion rate table fourth The dilation rate value of the layer dimension is sampled, and the fourth eigenvalue is obtained after the convolution calculation; the eigenvalue of the pixel of the output channel of the second layer of the dilated convolution layer is obtained according to the third eigenvalue and the fourth eigenvalue. The channel pixels of the first layer dimension of the feature map query the expansion rate table of the fifth layer dimension to sample the expansion rate value of the fifth layer dimension, and obtain the fifth feature value after convolution calculation; the channel pixels of the second layer dimension query the expansion rate table sixth The dilation rate value of the layer dimension is sampled, and the sixth eigenvalue is obtained after the convolution calculation; the eigenvalue of the pixel of the output channel of the third layer of the dilated convolution layer is obtained according to the fifth eigenvalue and the sixth eigenvalue. In this way, a feature map with three output channels of the dilated convolutional layer can be obtained.

In some exemplary embodiments of the present disclosure, the filters of each output channel of the dilated convolutional layer may also be grouped according to preset rules; The expansion rate table corresponding to each group performs expansion and convolution processing on the feature map; wherein, in each group, the same coordinate pixel corresponding to each output channel adopts the same expansion rate value.

For example, the filter bank corresponding to each output channel of the dilated convolution layer can be divided into several groups, and the same coordinate pixels of different output channels in the group adopt the same dilation rate. For example, if the input feature map image size is Width*height*channel_in, and the number of groups is group, then the image size of the expansion rate table is Width*height*(channel_in*group). In the dilated convolution layer, the filters can be divided into groups according to actual needs, such as 4, 6 or 8 groups. For example, when channel_in=2, group=3, and the number of output channels of the dilated convolutional layer is channel_out=3, the image size of the dilation rate table is Width*height*6. For example, when the dilated convolutional layer channel_out=32 and group=4, then the 1-4 layers use the same dilation rate table, the 5-8 layers use the same dilation rate table, and so on. By dividing the filters into groups, the complexity of the network can be reduced, computing resources can be saved, and the accuracy of the dilated convolution operation can be guaranteed.

In this example implementation, an image feature extraction model is provided, including a dilation rate calculation layer and a dilation convolution layer. The number of input channels for this model is 2 and the number of output channels is 3. The expansion rate calculation layer contains 1 convolutional layer, the filter size is 3*3, and the output channel is 2. This layer performs a convolution operation on the input feature map and outputs a dilation rate table of the same image size. The expansion convolution layer first looks up the expansion rate table to determine the expansion rate value corresponding to the pixel, expands the sampling range, and then calculates.

For example, for the pixel whose coordinates are the 100th row, 100th column, and the first channel in the input feature map, look up the same coordinates in the expansion rate array, and find that the corresponding expansion rate is 1.5. If the filter size of the dilated convolutional layer is 3*3, the normal sampling coordinates of the pixel are as follows, the first number in parentheses represents the row coordinate, and the second number represents the column coordinate:

(99, 99) (99, 100) (99, 101)

(100, 99) (100, 100) (100, 101)

(101, 99) (101, 100) (101, 101)

And after the expansion rate is set to 1.5, its sampling coordinates become:

(98.5, 98.5) (98.5, 100) (98.5, 101.5)

(100, 98.5) (100, 100) (100, 101.5)

(101.5, 98.5) (101.5, 100) (101.5, 101.5)

Since the sampling coordinates are decimals, the bilinear interpolation method is used, using the integer coordinate pixels in the 4*4 neighborhood to calculate the weight of each pixel according to the distance, and estimate the eigenvalues at the sampling coordinates. Taking the coordinate point (98.5, 101.5) as an example, the integer coordinates of the reference pixel are from the 97th row to the 100th row, and from the 100th column to the 103rd column, specifically:

(97, 100) (97, 101) (97, 102) (97, 103)

(98, 100) (98, 101) (98, 102) (98, 103)

(99, 100) (99, 101) (99, 102) (99, 103)

(100, 100) (100, 101) (100, 102) (100, 103)

Furthermore, after estimating the eigenvalues at the sampling points, the results can be calculated by convolution.

The data processing methods provided by the embodiments of the present disclosure can be applied to deep learning neural networks for image noise reduction, image recognition or image registration, as well as semantic segmentation, face recognition, image reconstruction and other fields. By using the dilation rate calculation layer to configure the corresponding dilation rate table for the feature image, the receptive field can be adaptively and finely adjusted at the pixel level in the dilated convolution layer according to the image features. And the amount of computation is relatively small, which helps to improve the performance of semantic segmentation, image reconstruction and other fields. Taking the field of image noise reduction as an example, for the block noise caused by image compression, the receptive field needs to be enlarged as much as possible to dilute the noise in large blocks; for the texture area with rich details, it is necessary to sample as densely as possible to preserve the details. These two tasks contradict each other, and the actual processing is difficult to balance, or large blocks of noise remain, or high-frequency details are erased. However, this scheme can calculate the expansion rate at the pixel level according to the image characteristics, and adjust the receptive field adaptively, which is helpful to solve such problems.

It should be noted that the above-mentioned drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not indicate or limit the chronological order of these processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, for example, in multiple modules.

Further, as shown in FIG. 5 , the embodiment of this example further provides a data processing apparatus 50 , including: an adaptive expansion rate table calculation module 501 and a dilation convolution processing module 502 .

in,

The adaptive expansion rate table calculation module 501 can be used to obtain the feature map to be processed, and input the feature map into the expansion rate calculation layer to obtain the expansion rate table corresponding to the feature map; wherein, the expansion rate table Including the expansion rate value corresponding to each pixel in the feature map.

The dilated convolution processing module 502 can be configured to input the dilation rate table into the dilated convolution layer, so that the dilated convolution layer performs dilated convolution processing on the feature map in combination with the dilation rate table, to obtain: Feature data corresponding to the feature map.

In an example of the present disclosure, the expansion rate calculation layer includes at least one convolution layer; the adaptive expansion rate table calculation module 501 may be configured to perform convolution processing on the feature map at least once to obtain The expansion rate table.

In an example of the present disclosure, the expansion rate table image size configuration module (not shown in the figure).

The expansion rate table image size configuration module may be configured to configure the image size of the expansion rate table according to the image size of the feature map; or,

Configure the image size of the dilation rate table in combination with the image size of the feature map and the number of output channels of the dilated convolutional layer; or

The image size of the dilation rate table is configured in combination with the image size of the feature map and the filter grouping information of the output channel of the dilated convolutional layer.

In an example of the present disclosure, the image size of the expansion rate table is the same as the image size of the feature map; or

The image size of the dilation rate table is the product of the image size of the feature map and the number of output channels of the dilated convolutional layer; or

The image size of the dilation rate table is the product of the image size of the feature map and the number of filter groups of the output channel number of the dilated convolution layer.

In an example of the present disclosure, the dilated convolution processing module 502 may include: a query unit and a sampling unit (not shown in the figure).

The query unit may be configured to query the expansion rate table to obtain the expansion rate value corresponding to each pixel in the feature map.

The sampling unit may be configured to determine the sampling range of the corresponding pixel according to the dilation rate value, so as to perform dilation convolution processing.

In an example of the present disclosure, the query unit may include: a first query unit (not shown in the figure).

The first query unit can be used to perform dilation convolution processing for each layer output channel of the dilated convolution layer, and the same coordinate pixel in each layer output channel adopts the same dilation rate value to perform dilation convolution processing; wherein, the dilation rate table The same size as the image of the feature map.

In an example of the present disclosure, the query unit may include: a second query unit (not shown in the figure).

The second query unit may be used to query the dilation rate table corresponding to the output channels of each layer of the dilated convolution layer, so as to be used for the feature map to perform the dilation convolution in combination with the dilation rate table corresponding to each of the output channels respectively. processing; wherein, the image size of the dilation rate table is the product of the image size of the feature map and the number of output channels of the dilated convolution layer.

In an example of the present disclosure, the query unit may include: a third query unit (not shown in the figure).

The second query unit can be used to group the filters of each output channel of the dilated convolution layer according to preset rules; query the dilation rate table corresponding to each group according to the grouping result, so as to The dilation rate table performs dilation and convolution processing on the feature map; wherein, in each group, the same coordinate pixel corresponding to each output channel adopts the same dilation rate value.

In an example of the present disclosure, the sampling unit may include: based on the dilation rate value corresponding to the pixel, using the same interval for sampling in the horizontal direction and the vertical direction of the pixel.

In an example of the present disclosure, the expansion ratio value is an integer, or a non-integer.

In an example of the present disclosure, the dilated convolution processing module 502 may further include: a neighborhood interpolation calculation unit. (not shown in the figure).

The neighborhood interpolation calculation unit may be configured to obtain neighborhood pixel values of the sampling points corresponding to the pixels when the expansion rate value is non-integer, and calculate the corresponding neighborhood pixel interpolation values based on the neighborhood pixel values ; Estimating the eigenvalues of the sampling points according to the neighborhood pixel interpolation, so as to perform dilated convolution based on the eigenvalues of the sampling points.

The specific details of each module in the above-mentioned data processing apparatus have been described in detail in the corresponding data processing method, and thus will not be repeated here.

It should be noted that although several modules or units of the apparatus for action performance are mentioned in the above detailed description, this division is not mandatory. Indeed, according to embodiments of the present disclosure, the features and functions of two or more modules or units described above may be embodied in one module or unit. Conversely, the features and functions of one module or unit described above may be further divided into multiple modules or units to be embodied.

Figure 6 shows a schematic diagram of a wireless communication device suitable for implementing embodiments of the present invention.

It should be noted that the electronic device 600 shown in FIG. 6 is only an example, and should not impose any limitations on the functions and scope of use of the embodiments of the present disclosure.

As shown in FIG. 6 , the electronic device 600 may specifically include: a processor 610, an internal memory 621, an external memory interface 622, a Universal Serial Bus (USB) interface 630, a charging management module 640, a power management module 641, Battery 642, Antenna 1, Antenna 2, Mobile Communication Module 650, Wireless Communication Module 660, Audio Module 670, Speaker 671, Receiver 672, Microphone 673, Headphone Interface 674, Sensor Module 680, Display Screen 690, Camera Module 691, Indication 692, a motor 693, a key 694, a subscriber identification module (SIM) card interface 695, and the like. The sensor module 680 may include a depth sensor 6801, a pressure sensor 6802, a gyroscope sensor 6803, an air pressure sensor 6804, a magnetic sensor 6805, an acceleration sensor 6806, a distance sensor 6807, a proximity light sensor 6808, a fingerprint sensor 6809, a temperature sensor 6810, and a touch sensor. 6811, ambient light sensor 6812, bone conduction sensor 6813, etc.

It can be understood that the structures illustrated in the embodiments of the present application do not constitute a specific limitation on the electronic device 600 . In other embodiments of the present application, the electronic device 600 may include more or less components than shown, or combine some components, or separate some components, or arrange different components. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

The processor 610 may include one or more processing units, for example, the processor 610 may include an application processor (Application Processor, AP), a modem processor, a graphics processor (Graphics Processing Unit, GPU), an image signal processor ( Image Signal Processor, ISP), controller, video codec, digital signal processor (Digital Signal Processor, DSP), baseband processor and/or neural network processor (Neural-network Processing Unit, NPU), etc. Wherein, different processing units may be independent devices, or may be integrated in one or more processors.

The controller can generate an operation control signal according to the instruction operation code and timing signal, and complete the control of fetching and executing instructions.

A memory may also be provided in the processor 610 for storing instructions and data. The memory may store instructions for implementing six modular functions: detection instructions, connection instructions, information management instructions, analysis instructions, data transmission instructions, and notification instructions, and the execution is controlled by the processor 610 . In some embodiments, the memory in processor 610 is cache memory. This memory may hold instructions or data that have just been used or recycled by the processor 610 . If the processor 610 needs to use the instruction or data again, it can be called directly from the memory. Repeated accesses are avoided and the waiting time of the processor 610 is reduced, thereby increasing the efficiency of the system.

In some embodiments, the processor 610 may include one or more interfaces. The interface may include an integrated circuit (Inter-Integrated Circuit, I2C) interface, an integrated circuit built-in audio (Inter-Integrated CircuitSound, I2S) interface, a pulse code modulation (Pulse Code Modulation, PCM) interface, a Universal Asynchronous Receiver (Universal Asynchronous Receiver) /Rransmitter, UART) interface, Mobile Industry Processor Interface (MIPI), General-PurposeInput/Output (GPIO) interface, Subscriber Identity Module (SIM) interface and/or general purpose Serial bus (Universal Serial Bus, USB) interface and so on.

The I2C interface is a bidirectional synchronous serial bus, including a serial data line (Serial Data line, SDA) and a serial clock line (Serail Clock line, SCL). In some embodiments, the processor 610 may contain multiple sets of I2C buses. The processor 610 can be respectively coupled to the touch sensor 6811, the charger, the flash, the camera module 691 and the like through different I2C bus interfaces. For example, the processor 610 may couple the touch sensor 6811 through the I2C interface, so that the processor 610 and the touch sensor 6811 communicate with each other through the I2C bus interface, so as to realize the touch function of the electronic device 600 .

The I2S interface can be used for audio communication. In some embodiments, the processor 610 may contain multiple sets of I2S buses. The processor 610 may be coupled with the audio module 670 through an I2S bus to implement communication between the processor 610 and the audio module 670 . In some embodiments, the audio module 670 can transmit audio signals to the wireless communication module 660 through the I2S interface, so as to realize the function of answering calls through a Bluetooth headset.

The PCM interface can also be used for audio communications, sampling, quantizing and encoding analog signals. In some embodiments, audio module 670 and wireless communication module 660 may be coupled through a PCM bus interface. In some embodiments, the audio module 670 can also transmit audio signals to the wireless communication module 660 through the PCM interface, so as to realize the function of answering calls through the Bluetooth headset. Both the I2S interface and the PCM interface can be used for audio communication.

The UART interface is a universal serial data bus used for asynchronous communication. The bus may be a bidirectional communication bus. It converts the data to be transmitted between serial communication and parallel communication. In some embodiments, a UART interface is typically used to connect the processor 610 with the wireless communication module 660 . For example, the processor 610 communicates with the Bluetooth module in the wireless communication module 660 through the UART interface to implement the Bluetooth function. In some embodiments, the audio module 670 can transmit audio signals to the wireless communication module 660 through the UART interface, so as to realize the function of playing music through the Bluetooth headset.

The MIPI interface can be used to connect the processor 610 with the display screen 690, the camera module 691 and other peripheral devices. The MIPI interface includes a camera serial interface (Camera Serial Interface, CSI), a display serial interface (DisplaySerial Interface, DSI), and the like. In some embodiments, the processor 610 communicates with the camera module 691 through a CSI interface to implement the photographing function of the electronic device 600 . The processor 610 communicates with the display screen 690 through the DSI interface to implement the display function of the electronic device 600 .

The GPIO interface can be configured by software. The GPIO interface can be configured as a control signal or as a data signal. In some embodiments, the GPIO interface may be used to connect the processor 610 with the camera module 691, the display screen 690, the wireless communication module 660, the audio module 670, the sensor module 680, and the like. The GPIO interface can also be configured as I2C interface, I2S interface, UART interface, MIPI interface, etc.

The USB interface 630 is an interface conforming to the USB standard specification, and may specifically be a MiniUSB interface, a MicroUSB interface, a USBTypeC interface, and the like. The USB interface 630 can be used to connect a charger to charge the electronic device 600, and can also be used to transmit data between the electronic device 600 and peripheral devices. It can also be used to connect headphones to play audio through the headphones. The interface can also be used to connect other electronic devices, such as AR devices.

It can be understood that, the interface connection relationship between the modules illustrated in the embodiment of the present invention is only a schematic illustration, and does not constitute a structural limitation of the electronic device 600 . In other embodiments of the present application, the electronic device 600 may also adopt different interface connection manners in the foregoing embodiments, or a combination of multiple interface connection manners.

The charging management module 640 is used to receive charging input from the charger. The charger may be a wireless charger or a wired charger. In some wired charging embodiments, the charging management module 640 may receive charging input from the wired charger through the USB interface 630 . In some wireless charging embodiments, the charging management module 640 may receive wireless charging input through a wireless charging coil of the electronic device 600 . While the charging management module 640 is charging the battery 642 , it can also supply power to the electronic device through the power management module 641 .

The power management module 641 is used for connecting the battery 642 , the charging management module 640 and the processor 610 . The power management module 641 receives input from the battery 642 and/or the charging management module 640, and supplies power to the processor 610, the internal memory 621, the display screen 690, the camera module 691, the wireless communication module 660, and the like. The power management module 641 can also be used to monitor parameters such as battery capacity, battery cycle times, battery health status (leakage, impedance). In some other embodiments, the power management module 641 may also be provided in the processor 610 . In other embodiments, the power management module 641 and the charging management module 640 may also be provided in the same device.

The wireless communication function of the electronic device 600 can be implemented by the antenna 1, the antenna 2, the mobile communication module 650, the wireless communication module 660, the modulation and demodulation processor, the baseband processor, and the like.

Antenna 1 and Antenna 2 are used to transmit and receive electromagnetic wave signals. Each antenna in electronic device 600 may be used to cover a single or multiple communication frequency bands. Different antennas can also be reused to improve antenna utilization. For example, the antenna 1 can be multiplexed as a diversity antenna of the wireless local area network. In other embodiments, the antenna may be used in conjunction with a tuning switch.

The mobile communication module 650 can provide a wireless communication solution including 2G/3G/4G/5G, etc. applied on the electronic device 600 . The mobile communication module 650 may include at least one filter, switch, power amplifier, low noise amplifier (Low Noise Amplifier, LNA) and the like. The mobile communication module 650 can receive electromagnetic waves from the antenna 1, filter and amplify the received electromagnetic waves, and transmit them to the modulation and demodulation processor for demodulation. The mobile communication module 650 can also amplify the signal modulated by the modulation and demodulation processor, and then convert it into electromagnetic waves for radiation through the antenna 1 . In some embodiments, at least part of the functional modules of the mobile communication module 650 may be provided in the processor 610 . In some embodiments, at least part of the functional modules of the mobile communication module 650 may be provided in the same device as at least part of the modules of the processor 610 .

The modem processor may include a modulator and a demodulator. Wherein, the modulator is used to modulate the low frequency baseband signal to be sent into a medium and high frequency signal. The demodulator is used to demodulate the received electromagnetic wave signal into a low frequency baseband signal. Then the demodulator transmits the demodulated low-frequency baseband signal to the baseband processor for processing. The low frequency baseband signal is processed by the baseband processor and passed to the application processor. The application processor outputs a sound signal through an audio device (not limited to the speaker 671 , the receiver 672 , etc.), or displays an image or video through the display screen 690 . In some embodiments, the modem processor may be a stand-alone device. In other embodiments, the modem processor may be independent of the processor 610, and may be provided in the same device as the mobile communication module 650 or other functional modules.

The wireless communication module 660 can provide wireless local area networks (Wireless Local Area Networks, WLAN) (such as Wireless Fidelity (Wi-Fi) networks), Bluetooth (Bluetooth, BT), and global navigation satellite systems applied on the electronic device 600. (Global Navigation Satellite System, GNSS), frequency modulation (Frequency Modulation, FM), near field communication technology (Near Field Communication, NFC), infrared technology (Infrared, IR) and other wireless communication solutions. The wireless communication module 660 may be one or more devices integrating at least one communication processing module. The wireless communication module 660 receives electromagnetic waves via the antenna 2 , frequency modulates and filters the electromagnetic wave signals, and sends the processed signals to the processor 610 . The wireless communication module 660 can also receive the signal to be sent from the processor 610 , perform frequency modulation on the signal, amplify the signal, and then convert it into an electromagnetic wave for radiation through the antenna 2 .

In some embodiments, the antenna 1 of the electronic device 600 is coupled with the mobile communication module 650, and the antenna 2 is coupled with the wireless communication module 660, so that the electronic device 600 can communicate with the network and other devices through wireless communication technology. The wireless communication technology may include Global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Wideband Code Division Multiple Access (WCDMA), Time-Division Code Division Multiple Access (TDSCDMA), Long Term Evolution (LTE), BT, GNSS, WLAN, NFC, FM, and / or IR technology, etc. The GNSS may include Global Positioning System (GPS), Global Navigation Satellite System (GLONASS), Beidou satellite navigation system (Beidouavigation satellite system, BDS), Quasi-Zenith satellite system (Quasi- Zenith Satellite System, QZSS) and/or Satellite Based Augmentation Systems (SBAS).

The electronic device 600 implements a display function through a GPU, a display screen 690, an application processor, and the like. The GPU is a microprocessor for image processing, and is connected to the display screen 690 and the application processor. The GPU is used to perform mathematical and geometric calculations for graphics rendering. Processor 610 may include one or more GPUs that execute program instructions to generate or alter display information.

Display screen 690 is used to display images, videos, and the like. Display screen 690 includes a display panel. The display panel can be a Liquid Crystal Display (LCD), an Organic Light Emitting Diode (OLED), an Active Matrix Organic Light Emitting Diode or an Active Matrix Organic Light Emitting Diode (Active-Matrix Organic Light Emitting Diode). , AMOLED), flexible light-emitting diode (Flexlight-Emitting Diode, FLED), Miniled, MicroLed, Micro-oLed, quantum dot light-emitting diode (Quantum dot Light Emitting Diodes, QLED) and so on. In some embodiments, the electronic device 600 may include 1 or N display screens 690 , where N is a positive integer greater than 1.

The electronic device 600 can realize the shooting function through the ISP, the camera module 691, the video codec, the GPU, the display screen 690, the application processor, and the like.

The ISP is used to process the data fed back by the camera module 691 . For example, when taking a photo, the shutter is opened, the light is transmitted to the camera photosensitive element through the lens, the light signal is converted into an electrical signal, and the camera photosensitive element transmits the electrical signal to the ISP for processing, and converts it into an image visible to the naked eye. ISP can also perform algorithm optimization on image noise, brightness, and skin tone. ISP can also optimize the exposure, color temperature and other parameters of the shooting scene. In some embodiments, the ISP may be provided in the camera module 691 .

The camera module 691 is used to capture still images or videos. The object is projected through the lens to generate an optical image onto the photosensitive element. The photosensitive element can be a charge coupled device (Charge Coupled Device, CCD) or a complementary metal-oxide-semiconductor (Complementary Metal-Oxide-Semiconductor, CMOS) phototransistor. The photosensitive element converts the optical signal into an electrical signal, and then transmits the electrical signal to the ISP to convert it into a digital image signal. The ISP outputs the digital image signal to the DSP for processing. DSP converts digital image signals into standard RGB, YUV and other formats of image signals. In some embodiments, the electronic device 600 may include 1 or N camera modules 691, where N is a positive integer greater than 1. If the electronic device 600 includes N cameras, one of the N cameras is the main camera.

A digital signal processor is used to process digital signals, in addition to processing digital image signals, it can also process other digital signals. For example, when the electronic device 600 selects a frequency point, the digital signal processor is used to perform Fourier transform on the frequency point energy, and the like.

Video codecs are used to compress or decompress digital video. Electronic device 600 may support one or more video codecs. In this way, the electronic device 600 can play or record videos of various encoding formats, such as: Moving Picture Experts Group (Moving Picture Experts Group, MPEG) 1, MPEG2, MPEG3, MPEG4, and so on.

NPU is a neural network (Neural-Network, NN) computing processor. By borrowing the structure of biological neural network, such as the transmission mode between neurons in the human brain, it can quickly process the input information and can continuously learn by itself. Applications such as intelligent cognition of the electronic device 600 can be implemented through the NPU, such as image recognition, face recognition, speech recognition, text understanding, and the like.

The external memory interface 622 can be used to connect an external memory card, such as a Micro SD card, to expand the storage capacity of the electronic device 600 . The external memory card communicates with the processor 610 through the external memory interface 622 to realize the data storage function. For example to save files like music, video etc in external memory card.

Internal memory 621 may be used to store computer executable program code, which includes instructions. The internal memory 621 may include a storage program area and a storage data area. The storage program area can store an operating system, an application program required for at least one function (such as a sound playback function, an image playback function, etc.), and the like. The storage data area may store data (such as audio data, phone book, etc.) created during the use of the electronic device 600 and the like. In addition, the internal memory 621 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, Universal Flash Storage (UFS), and the like. The processor 610 executes various functional applications and data processing of the electronic device 600 by executing instructions stored in the internal memory 621 and/or instructions stored in a memory provided in the processor.

The electronic device 600 may implement audio functions through an audio module 670, a speaker 671, a receiver 672, a microphone 673, an earphone interface 674, an application processor, and the like. Such as music playback, recording, etc.

The audio module 670 is used for converting digital audio information into analog audio signal output, and also for converting analog audio input into digital audio signal. Audio module 670 may also be used to encode and decode audio signals. In some embodiments, the audio module 670 may be provided in the processor 610 , or some functional modules of the audio module 670 may be provided in the processor 610 .

The speaker 671, also called "speaker", is used to convert audio electrical signals into sound signals. The electronic device 600 can listen to music through the speaker 671, or listen to a hands-free call.

The receiver 672, also referred to as an "earpiece", is used to convert audio electrical signals into sound signals. When the electronic device 600 answers a call or a voice message, the voice can be answered by placing the receiver 672 close to the human ear.

The microphone 673, also called "microphone" or "microphone", is used to convert sound signals into electrical signals. When making a call or sending a voice message, the user can make a sound by approaching the microphone 673 through the human mouth, and input the sound signal into the microphone 673 . The electronic device 600 may be provided with at least one microphone 673 . In other embodiments, the electronic device 600 may be provided with two microphones 673, which can implement a noise reduction function in addition to collecting sound signals. In other embodiments, the electronic device 600 may further be provided with three, four or more microphones 673 to collect sound signals, reduce noise, identify sound sources, and implement directional recording functions.

The earphone jack 674 is used to connect wired earphones. The earphone interface 674 may be the USB interface 630, or may be a 3.5mm Open Mobile Terminal Platform (OMTP) standard interface, a Cellular Telecommunications Industry Association of the USA (CTIA) standard interface.

The depth sensor 6801 is used to acquire depth information of the scene. In some embodiments, the depth sensor may be disposed in the camera module 691 .

The pressure sensor 6802 is used to sense pressure signals, and can convert the pressure signals into electrical signals. In some embodiments, the pressure sensor 6802 may be provided on the display screen 690 . There are many types of pressure sensors 6802, such as resistive pressure sensors, inductive pressure sensors, capacitive pressure sensors, and the like. The capacitive pressure sensor may be comprised of at least two parallel plates of conductive material. When a force is applied to the pressure sensor 6802, the capacitance between the electrodes changes. The electronic device 600 determines the intensity of the pressure according to the change in capacitance. When a touch operation acts on the display screen 690 , the electronic device 600 detects the intensity of the touch operation according to the pressure sensor 6802 . The electronic device 600 can also calculate the touched position according to the detection signal of the pressure sensor 6802 . In some embodiments, touch operations acting on the same touch position but with different touch operation intensities may correspond to different operation instructions. For example, when a touch operation whose intensity is less than the first pressure threshold acts on the short message application icon, the instruction for viewing the short message is executed. When a touch operation with a touch operation intensity greater than or equal to the first pressure threshold acts on the short message application icon, the instruction to create a new short message is executed.

The gyro sensor 6803 can be used to determine the motion attitude of the electronic device 600 . In some embodiments, the angular velocity of the electronic device 600 about three axes (ie, the x, y, and z axes) may be determined by the gyro sensor 6803 . The gyro sensor 6803 can be used for image stabilization. Exemplarily, when the shutter is pressed, the gyro sensor 6803 detects the angle at which the electronic device 600 shakes, calculates the distance that the lens module needs to compensate for according to the angle, and allows the lens to counteract the shake of the electronic device 600 through reverse motion to achieve anti-shake. The gyroscope sensor 6803 can also be used for navigation and somatosensory game scenes.

Air pressure sensor 6804 is used to measure air pressure. In some embodiments, the electronic device 600 uses the air pressure value measured by the air pressure sensor 6804 to calculate the altitude to aid in positioning and navigation.

The magnetic sensor 6805 includes a Hall sensor. The electronic device 600 can use the magnetic sensor 6805 to detect the opening and closing of the flip holster. In some embodiments, when the electronic device 600 is a flip machine, the electronic device 600 can detect the opening and closing of the flip according to the magnetic sensor 6805 . Further, according to the detected opening and closing state of the leather case or the opening and closing state of the flip cover, characteristics such as automatic unlocking of the flip cover are set.

The acceleration sensor 6806 can detect the magnitude of the acceleration of the electronic device 600 in various directions (generally three axes). The magnitude and direction of gravity can be detected when the electronic device 600 is stationary. It can also be used to identify the posture of electronic devices, and can be used in applications such as horizontal and vertical screen switching, pedometers, etc.

Distance sensor 6807 for measuring distance. The electronic device 600 can measure the distance through infrared or laser. In some embodiments, when shooting a scene, the electronic device 600 can use the distance sensor 6807 to measure the distance to achieve fast focusing.

Proximity light sensor 6808 may include, for example, light emitting diodes (LEDs) and light detectors, such as photodiodes. The light emitting diodes may be infrared light emitting diodes. The electronic device 600 emits infrared light to the outside through light emitting diodes. Electronic device 600 uses photodiodes to detect infrared reflected light from nearby objects. When sufficient reflected light is detected, it may be determined that there is an object near the electronic device 600 . When insufficient reflected light is detected, the electronic device 600 may determine that there is no object near the electronic device 600 . The electronic device 600 can use the proximity light sensor 6808 to detect that the user holds the electronic device 600 close to the ear to talk, so as to automatically turn off the screen to save power. Proximity light sensor 6808 can also be used in holster mode, pocket mode automatically unlock and lock screen.

The fingerprint sensor 6809 is used to collect fingerprints. The electronic device 600 can use the collected fingerprint characteristics to realize fingerprint unlocking, accessing application locks, taking photos with fingerprints, answering incoming calls with fingerprints, and the like.

The temperature sensor 6810 is used to detect the temperature. In some embodiments, the electronic device 600 utilizes the temperature detected by the temperature sensor 6810 to execute a temperature processing strategy. For example, when the temperature reported by the temperature sensor 6810 exceeds a threshold, the electronic device 600 performs a performance reduction of the processor located near the temperature sensor 6810 in order to reduce power consumption and implement thermal protection. In other embodiments, when the temperature is lower than another threshold, the electronic device 600 heats the battery 642 to avoid abnormal shutdown of the electronic device 600 caused by the low temperature. In some other embodiments, when the temperature is lower than another threshold, the electronic device 600 boosts the output voltage of the battery 642 to avoid abnormal shutdown caused by low temperature.

The touch sensor 6811 is also called "touch device". The touch sensor 6811 can be disposed on the display screen 690, and the touch sensor 6811 and the display screen 690 form a touch screen, also called a "touch screen". The touch sensor 6811 is used to detect a touch operation on or near it. The touch sensor can pass the detected touch operation to the application processor to determine the type of touch event. Visual output related to touch operations may be provided through display screen 690 . In other embodiments, the touch sensor 6811 may also be disposed on the surface of the electronic device 600 , which is different from the position where the display screen 690 is located.

The ambient light sensor 6812 is used to sense ambient light brightness. The electronic device 600 can adaptively adjust the brightness of the display screen 690 according to the perceived ambient light brightness. The ambient light sensor 6812 can also be used to automatically adjust the white balance when taking pictures. The ambient light sensor 6812 can also cooperate with the proximity light sensor 6808 to detect whether the electronic device 600 is in a pocket to prevent accidental touches.

The bone conduction sensor 6813 can acquire vibration signals. In some embodiments, the bone conduction sensor 6813 can acquire the vibration signal of the vibrating bone mass of the human voice. The bone conduction sensor 6813 can also contact the pulse of the human body and receive the blood pressure beating signal. In some embodiments, the bone conduction sensor 6813 can also be disposed in the earphone, combined with the bone conduction earphone. The audio module 670 can parse out the voice signal based on the vibration signal of the voice part vibrating the bone mass obtained by the bone conduction sensor 6813, so as to realize the voice function. The application processor can analyze the heart rate information based on the blood pressure beat signal obtained by the bone conduction sensor 6813 to realize the heart rate detection function.

The keys 694 include a power-on key, a volume key, and the like. Keys 694 may be mechanical keys. It can also be a touch key. The electronic device 600 may receive key inputs and generate key signal inputs related to user settings and function control of the electronic device 600 .

Motor 693 can generate vibrating cues. The motor 693 can be used for incoming call vibration alerts, and can also be used for touch vibration feedback. For example, touch operations acting on different applications (such as taking pictures, playing audio, etc.) can correspond to different vibration feedback effects. The motor 693 can also correspond to different vibration feedback effects for touch operations on different areas of the display screen 690 . Different application scenarios (for example: time reminder, receiving information, alarm clock, games, etc.) can also correspond to different vibration feedback effects. The touch vibration feedback effect can also support customization.

The indicator 692 can be an indicator light, which can be used to indicate the charging status, the change of power, and can also be used to indicate messages, missed calls, notifications, and the like.

The SIM card interface 695 is used to connect a SIM card. The SIM card can be contacted and separated from the electronic device 600 by inserting into the SIM card interface 695 or pulling out from the SIM card interface 695 . The electronic device 600 may support 1 or N SIM card interfaces, where N is a positive integer greater than 1. The SIM card interface 695 can support Nano SIM card, Micro SIM card, SIM card and so on. The same SIM card interface 695 can insert multiple cards at the same time. The types of the plurality of cards may be the same or different. The SIM card interface 695 can also be compatible with different types of SIM cards. The SIM card interface 695 is also compatible with external memory cards. The electronic device 600 interacts with the network through the SIM card to implement functions such as call and data communication. In some embodiments, the electronic device 600 employs an eSIM, ie: an embedded SIM card. The eSIM card can be embedded in the electronic device 600 and cannot be separated from the electronic device 600 .

In particular, the processes described below with reference to the flowcharts may be implemented as computer software programs according to embodiments of the present invention. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the method illustrated in the flowchart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication portion, and/or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), various functions defined in the system of the present application are performed.

It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or a combination of any of the above. More specific examples of computer readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable Erasable Programmable Read Only Memory (EPROM), flash memory, optical fiber, portable Compact Disc Read-Only Memory (CD-ROM), optical storage device, magnetic storage device, or any suitable of the above The combination. In the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present invention, however, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code therein. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device . Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination of the foregoing.

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code that contains one or more logical functions for implementing the specified functions executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It is also noted that each block of the block diagrams or flowchart illustrations, and combinations of blocks in the block diagrams or flowchart illustrations, can be implemented in special purpose hardware-based systems that perform the specified functions or operations, or can be implemented using A combination of dedicated hardware and computer instructions is implemented.

The units involved in the embodiments of the present invention may be implemented in a software manner, or may be implemented in a hardware manner, and the described units may also be provided in a processor. Among them, the names of these units do not constitute a limitation on the unit itself under certain circumstances.

It should be noted that, as another aspect, the present application also provides a computer-readable medium, and the computer-readable medium may be included in the electronic device described in the foregoing embodiments; assembled into the electronic device. The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by an electronic device, causes the electronic device to implement the methods described in the following embodiments. For example, the electronic device can implement the various steps shown in FIG. 2 .

Furthermore, the above-mentioned figures are merely schematic illustrations of the processes included in the methods according to the exemplary embodiments of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not indicate or limit the chronological order of these processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, for example, in multiple modules.

Other embodiments of the present disclosure will readily suggest themselves to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or techniques in the technical field not disclosed by the present disclosure . The specification and examples are to be regarded as exemplary only, with the true scope and spirit of the disclosure being indicated by the claims.

It is to be understood that the present disclosure is not limited to the precise structures described above and illustrated in the accompanying drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims (14)

1. a data processing method, is characterized in that, comprises: Obtain the feature map to be processed, and input the feature map into the expansion rate calculation layer to obtain the expansion rate table corresponding to the feature map; wherein, the expansion rate table includes the expansion rate value corresponding to each pixel in the feature map ; The dilation rate table is input into the dilation convolution layer, so that the dilation convolution layer performs dilation convolution processing on the feature map in combination with the dilation rate table, so as to obtain feature data corresponding to the feature map. 2. The data processing method according to claim 1, wherein the expansion rate calculation layer comprises at least one convolution layer; The inputting the feature map into the expansion rate calculation layer to obtain the expansion rate table corresponding to the feature map, including: Perform at least one convolution process on the feature map to obtain the dilation rate table. 3. The data processing method according to claim 1 or 2, wherein the method further comprises: Configure the image size of the expansion rate table according to the image size of the feature map; or, Configure the image size of the dilation rate table in combination with the image size of the feature map and the number of output channels of the dilated convolutional layer; or The image size of the dilation rate table is configured in combination with the image size of the feature map and the filter grouping information of the output channel of the dilated convolutional layer. 4. The data processing method according to claim 3, wherein the image size of the expansion rate table is the same as the image size of the feature map; or The image size of the dilation rate table is the product of the image size of the feature map and the number of output channels of the dilated convolutional layer; or The image size of the dilation rate table is the product of the image size of the feature map and the number of filter groups of the output channel number of the dilated convolution layer. 5 . The data processing method according to claim 1 , wherein the dilation rate table is input into a dilated convolution layer, so that the dilated convolution layer combines the dilation rate table with the feature map. 6 . Perform dilated convolution processing, including: query the expansion rate table to obtain the expansion rate value corresponding to each pixel in the feature map; The sampling range of the corresponding pixel is determined according to the dilation rate value, so as to perform dilation convolution processing. 6. The data processing method according to claim 5, wherein the querying the expansion rate table to obtain the expansion rate value corresponding to each pixel in the feature map comprises: For the output channels of each layer of the dilated convolution layer, the same coordinate pixel adopts the same dilation rate value in the output channels of each layer for dilation convolution processing; wherein, the dilation rate table is the same as the image size of the feature map . 7. The data processing method according to claim 5, wherein the querying the expansion rate table to obtain the expansion rate value corresponding to each pixel in the feature map comprises: Query the dilation rate table corresponding to the output channels of each layer of the dilated convolution layer, so as to use the feature map to perform dilation and convolution processing in combination with the dilation rate table corresponding to each of the output channels; wherein, the dilation rate table The image size of is the product of the feature map image size and the number of output channels of the dilated convolutional layer. 8. The data processing method according to claim 5, wherein the querying the expansion rate table to obtain the expansion rate value corresponding to each pixel in the feature map comprises: Group the filters of each output channel of the dilated convolutional layer according to preset rules; Query the expansion rate table corresponding to each group according to the grouping result, so as to perform expansion and convolution processing on the feature map according to the expansion rate table corresponding to each group; wherein, in each group, the same coordinate pixel corresponding to each output channel adopts the same expansion rate value. 9. The data processing method according to any one of claims 5-8, wherein the determining the sampling range of the corresponding pixel according to the expansion rate value comprises: Based on the dilation rate value corresponding to the pixel, sampling is performed using the same interval in the horizontal direction and the vertical direction of the pixel. 10 . The data processing method according to claim 5 , wherein the expansion rate value is an integer or a non-integer. 11 . 11 . The data processing method according to claim 10 , wherein when the dilation rate value is non-integer, the performing dilation convolution according to the dilation rate value corresponding to each of the pixels, comprising: 11 . : Obtain the neighborhood pixel value of the sampling point corresponding to the pixel, and calculate the corresponding neighborhood pixel interpolation based on the neighborhood pixel value; The eigenvalues of the sampling points are estimated according to the neighborhood pixel interpolation to perform dilated convolution based on the eigenvalues of the sampling points. 12. A data processing device, comprising: The adaptive expansion rate table calculation module is used to obtain the feature map to be processed, and input the feature map into the expansion rate calculation layer to obtain the expansion rate table corresponding to the feature map; wherein, the expansion rate table includes the The expansion rate value corresponding to each pixel in the feature map; a dilated convolution processing module, configured to input the dilation rate table into the dilated convolution layer, so that the dilated convolution layer combines the dilation rate table to perform dilated convolution processing on the feature map to obtain the feature The feature data corresponding to the graph. 13. A computer-readable medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the data processing method according to any one of claims 1 to 11 is implemented. 14. An electronic device, characterized in that, comprising: one or more processors; A storage device for storing one or more programs that, when executed by the one or more processors, cause the one or more processors to implement any one of claims 1 to 11 A method of data processing described.

Images