Model training method, display compensation method and model training apparatus

Abstract

The present disclosure belongs to the technical field of display. Provided are a model training method, a model training apparatus, a computer device and a computer-readable storage medium. The model training method of the present disclosure comprises: acquiring a first training sample set, wherein the first training sample set comprises a plurality of first sample groups, each first sample group comprises image information of a standard sub-pixel and image information of a sub-pixel to be subjected to compensation, the standard sub-pixel is a sub-pixel that is written with the same grayscale value as a sub-pixel co-powered therewith, and the sub-pixel to be subjected to compensation is a sub-pixel that is co-powered with a sub-pixel having a different color therefrom; using the first sample groups as an input of a first initial model to be trained, so as to obtain brightness compensation values of the sub-pixels to be subjected to compensation that are output by the first initial model; and on the basis of the brightness compensation values of the sub-pixels to be subjected to compensation and a preset standard brightness compensation value, training the first initial model, so as to obtain a brightness compensation model.

Description

Model training method, explicit compensation method, and model training device Technical Field

[0001] This disclosure belongs to the field of display technology, specifically relating to a model training method, a display compensation method, a model training device, a computer device, and a computer-readable storage medium. Background Technology

[0002] When display devices are manufactured, due to various reasons (such as circuitry and power supply voltage), there will always be errors that cause uneven brightness and textures, such as horizontal and vertical lines, to appear on the display screen. As shown in Figure 1, the circled thin lines represent extra bright elements, and it is these bright and dark pixels that cause the horizontal or vertical lines on the display screen. Summary of the Invention

[0003] The present invention aims to at least solve one of the technical problems existing in the prior art, and to provide a model training method, a display compensation method, a model training device, a computer device, and a computer-readable storage medium.

[0004] This disclosure provides a model training method, which includes:

[0005] Obtain a first training sample set; the first training sample set includes multiple first sample groups; the first sample group includes image information of standard sub-pixels and image information of sub-pixels to be compensated; the standard sub-pixel is a sub-pixel that shares the same grayscale value as a sub-pixel with the same electrical charge; the sub-pixel to be compensated is a sub-pixel that shares the same electrical charge as a sub-pixel with a different color.

[0006] The first sample group is used as the input to the first initial model to be trained, and the brightness compensation value of the sub-pixel to be compensated is obtained from the output of the first initial model.

[0007] Based on the brightness compensation value of the sub-pixel to be compensated and the preset standard brightness compensation value, the first initial model is trained to obtain a brightness compensation model.

[0008] The first initial model includes: a first feature extraction network, a second feature extraction network, and an output network;

[0009] The step of using the sample group as input to the first initial model to be trained, and obtaining the brightness compensation value of the sub-pixel to be compensated output by the first initial model, includes:

[0010] Using the image information of the standard sub-pixels in the sample group as the first feature extraction network, the image features of the standard sub-pixels output by the prime number first feature extraction network are obtained.

[0011] The image information of the sub-pixels to be compensated in the sample group is used as the second feature extraction network to obtain the image features of the sub-pixels to be compensated output by the prime number second feature extraction network.

[0012] The image features of the standard sub-pixel and the image features of the sub-pixel to be compensated are input into the output network to regress the brightness compensation value of the sub-pixel to be compensated.

[0013] The output network includes a splicing layer, a first fully connected layer, and a second fully connected layer.

[0014] The step of inputting the image features of the standard sub-pixel and the image features of the sub-pixel to be compensated into the output network to regress the brightness compensation value of the sub-pixel to be compensated includes:

[0015] The image features of the standard sub-pixel and the image features of the sub-pixel to be compensated are input into the stitching layer for feature stitching to obtain stitched features;

[0016] The spliced ​​features are input into the first fully connected layer for feature calculation to obtain the first intermediate features;

[0017] The first intermediate feature is input into the second fully connected layer for feature calculation, and the brightness compensation value of the sub-pixel to be compensated is regressed.

[0018] The step of obtaining the first training sample set includes:

[0019] Based on the connection relationship of the driving circuits of each sub-pixel in the display panel, multiple sampling areas of the display panel, as well as the position information of the standard sub-pixel and the sub-pixel to be compensated in the sampling areas, are determined and stored.

[0020] The control display panel displays according to a preset lighting mode, acquires the display image of the sampling area in the display panel, and extracts the image information of the standard sub-pixel and the image information of the sub-pixel to be compensated based on the stored position information of the standard sub-pixel and the sub-pixel to be compensated to obtain the first sample group. Multiple first sample groups constitute the first training sample set.

[0021] The sub-pixels of the display panel include sub-pixels of a first color, a second color, and a third color;

[0022] In some of the sample groups, the standard sub-pixels and the sub-pixels to be compensated satisfy the following conditions: the standard sub-pixel and the sub-pixel sharing the same electrical charge have the same color; the color of the standard sub-pixel and the sub-pixel to be compensated is a first color; and the sub-pixel to be compensated shares the same electrical charge with a sub-pixel of a third color.

[0023] In some of the sample groups, the standard sub-pixels and the sub-pixels to be compensated satisfy the following conditions: the standard sub-pixel and the sub-pixel sharing the same electrical charge have different colors, and the grayscale values ​​written to them are the same; the standard sub-pixel is a second color, and the sub-pixel sharing the same electrical charge with the standard sub-pixel is a first color; the colors of the sub-pixels to be compensated are all third colors; and the sub-pixels to be compensated share the same electrical charge with the sub-pixels of the second color.

[0024] In some of the sample groups, the standard sub-pixels and the sub-pixels to be compensated satisfy the following conditions: the standard sub-pixels are of a first color, and the sub-pixels to be compensated are of a second color; the sub-pixels to be compensated and the standard sub-pixels share the same electrical charge.

[0025] In some of the sample groups, the standard sub-pixels and the sub-pixels to be compensated satisfy the following conditions: the standard sub-pixel and the sub-pixel sharing the same electrical charge have the same color; the color of the standard sub-pixel and the sub-pixel to be compensated is a third color; and the sub-pixel to be compensated shares the same electrical charge with the sub-pixel of the first color.

[0026] Wherein, when the standard sub-pixel and the sub-pixel to be compensated in the sample group satisfy the condition that the standard sub-pixel and the sub-pixel sharing the same electrical charge have the same color, the color of the standard sub-pixel and the sub-pixel to be compensated are both the first color; when the sub-pixel to be compensated shares the electrical charge with the sub-pixel of the third color, the control display panel displays according to a preset lighting mode, acquires the display image of the sampling area in the display panel, and extracts the image information of the standard sub-pixel and the sub-pixel to be compensated based on the stored position information of the standard sub-pixel and the sub-pixel to be compensated to obtain the first sample group, the steps include:

[0027] The first color sub-pixels in the control display panel are written with the same first preset grayscale value, and the second color sub-pixels and the third color sub-pixels are written with the same second preset grayscale value. The display image of the sampling area in the display panel is obtained, and the image information of the standard sub-pixels and the image information of the sub-pixels to be compensated are extracted according to the stored position information of the standard sub-pixels and the sub-pixels to be compensated, so as to obtain the first sample group.

[0028] The steps of obtaining the first sample group include: when the standard sub-pixel and the sub-pixel to be compensated in the sample group satisfy the following conditions: the standard sub-pixel is a first color, and the sub-pixel to be compensated is a second color; when the sub-pixel to be compensated and the standard sub-pixel share the same electrical charge, the control display panel displays according to a preset lighting mode, acquires the display image of the sampling area in the display panel, and extracts the image information of the standard sub-pixel and the sub-pixel to be compensated based on the stored position information of the standard sub-pixel and the sub-pixel to be compensated.

[0029] The sub-pixels of the third color in the display panel are controlled to write 0 grayscale, and the sub-pixels of the first color and the second color are controlled to write the same preset grayscale value. The display image of the sampling area in the display panel is obtained, and the image information of the standard sub-pixel and the sub-pixel to be compensated is extracted according to the stored position information of the standard sub-pixel and the sub-pixel to be compensated, so as to obtain the first sample group.

[0030] The standard sub-pixel and the sub-pixel to be compensated in the sample group satisfy the following conditions: the standard sub-pixel and the sub-pixel sharing the same electrical charge have different colors, and both are written with the same grayscale value; the standard sub-pixel is a second color, and the sub-pixel sharing the same electrical charge is a first color; the color of the sub-pixel to be compensated is a third color; when the sub-pixel to be compensated shares the same electrical charge with the sub-pixel of the second color, the control display panel displays according to a preset lighting mode, acquires the display image of the sampling area in the display panel, and extracts the image information of the standard sub-pixel and the sub-pixel to be compensated based on the stored position information of the standard sub-pixel and the sub-pixel to be compensated to obtain the first sample group.

[0031] The first color subpixel and the second color subpixel in the control display panel are written with the same first preset grayscale value, and the third color subpixel is written with the same second preset grayscale value. The display image of the sampling area in the display panel is obtained, and the image information of the standard subpixel and the image information of the subpixel to be compensated are extracted according to the stored position information of the standard subpixel and the subpixel to be compensated to obtain the first sample group.

[0032] The steps for obtaining the first sample group include: when the standard sub-pixel and the sub-pixel to be compensated in the sample group satisfy the following conditions: the standard sub-pixel and the sub-pixel sharing the same electrical charge have the same color; the standard sub-pixel and the sub-pixel to be compensated are both a third color; and the sub-pixel to be compensated shares the same electrical charge with a sub-pixel of the first color, the control display panel displays the image according to a preset lighting mode, acquires the display image of the sampling area in the display panel, and extracts the image information of the standard sub-pixel and the sub-pixel to be compensated based on the stored position information of the standard sub-pixel and the sub-pixel to be compensated.

[0033] The sub-pixels of the third color in the display panel are controlled to display the same first preset grayscale value, and the sub-pixels of the first color are controlled to display the same second preset grayscale value. The display image of the sampling area in the display panel is obtained, and the image information of the standard sub-pixel and the sub-pixel to be compensated is extracted according to the stored position information of the standard sub-pixel and the sub-pixel to be compensated, so as to obtain the first sample group.

[0034] Wherein, the ratio of the image width of the standard sub-pixel image information to the width of the sub-pixel is 0.6 to 0.8; the ratio of the image length of the standard sub-pixel image information to the length of the sub-pixel is 0.6 to 0.8; the ratio of the image width of the sub-pixel image information to the width of the sub-pixel is 0.6 to 0.8; and the ratio of the image length of the sub-pixel image information to the length of the sub-pixel is 0.6 to 0.8.

[0035] The control display panel displays according to a preset lighting mode, and while acquiring the display image of the sampling area in the display panel, it also includes acquiring the stripe pattern of the display image of the sampling area; the model training method further includes:

[0036] A second sample group is generated based on the displayed image of the sampling area and the corresponding stripe pattern, and multiple second sample groups constitute a second training sample set;

[0037] The displayed image in the second sample group is used as the input of the second initial model to obtain the predicted stripe pattern output by the second initial model. Based on the predicted stripe pattern and the stripe pattern in the second sample group, the second model is trained to obtain a classification model.

[0038] The second initial model includes a third feature extraction network, a fourth feature extraction network, a fifth feature extraction network, a third fully connected layer, a fourth fully connected layer, and a classification layer connected in sequence. The steps of using the displayed images in the second sample group as input to the second initial model to obtain the predicted stripe pattern output by the second initial model, and training the second model based on the predicted stripe pattern and the stripe pattern in the second sample group to obtain a classification model, include:

[0039] The display image of the sampling area is input into the third feature extraction network, and features are extracted through the third feature extraction network, the fourth feature extraction network, and the fifth feature extraction network.

[0040] The output features of the fifth feature extraction network are processed by the third fully connected layer to obtain the second intermediate feature.

[0041] The third intermediate feature is obtained by performing feature operations on the second intermediate feature through the fourth fully connected layer;

[0042] The classification layer predicts the stripe pattern of the displayed image of the sampling area based on the third intermediate feature input, and obtains the predicted stripe pattern. Based on the predicted stripe pattern and the stripe pattern in the second sample group, the second model is trained to obtain the classification model.

[0043] The step of generating a second sample group based on the displayed image of the sampling area and the corresponding stripe pattern includes:

[0044] The display image of the sampling area is subjected to at least data processing to obtain a processed display image, and the processed display image and the corresponding stripe pattern are used as a second sample group.

[0045] The step of training the first initial model based on the brightness compensation value of the sub-pixel to be compensated and a preset standard brightness compensation value to obtain a brightness compensation model includes:

[0046] Based on the brightness compensation value of the sub-pixel to be compensated and the preset standard brightness compensation value, and using a loss function, the first initial model is trained. The model parameters of the first initial model are adjusted according to the output of the loss function until the loss function converges, thus obtaining the brightness compensation model.

[0047] This disclosure provides a display compensation method, which includes:

[0048] A brightness compensation table for each color channel of the display panel generated using a brightness compensation model; the brightness compensation model is obtained by training any of the model training methods described above.

[0049] The display data of the sub-pixels in the display panel pair are compensated based on the brightness compensation table.

[0050] Based on the compensated display data, the display panel controls the display of each sub-pixel.

[0051] This disclosure provides a model training apparatus, comprising:

[0052] The sample acquisition module is configured to acquire a first training sample set; the first training sample set includes multiple first sample groups; the first sample group includes image information of standard sub-pixels and image information of sub-pixels to be compensated; the standard sub-pixel is a sub-pixel that shares the same grayscale value as a sub-pixel with the same electrical charge; the sub-pixel to be compensated is a sub-pixel that shares the same electrical charge as a sub-pixel with a different color.

[0053] The training module is configured to take the sample group as input to the first initial model to be trained, obtain the brightness compensation value of the sub-pixel to be compensated output by the first initial model, and train the first initial model based on the brightness compensation value of the sub-pixel to be compensated and the preset standard brightness compensation value to obtain a brightness compensation model.

[0054] This disclosure provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements any of the methods described above.

[0055] This disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the methods described above. Attached Figure Description

[0056] Figure 1 shows a schematic diagram of the stripes appearing on the display panel.

[0057] Figure 2 is a schematic diagram of an exemplary display panel.

[0058] Figure 3 is a schematic diagram of the driving circuit connection of a local sub-pixel in the display panel.

[0059] Figure 4 is a flowchart of the model training method according to an embodiment of this disclosure.

[0060] Figure 5 is a flowchart of step S1 of the model training method according to an embodiment of the present disclosure.

[0061] Figure 6 is a schematic diagram of the position information of the standard sub-pixels and the position information of the sub-pixels to be compensated in the sampling area of ​​an embodiment of this disclosure.

[0062] Figure 7 is a schematic diagram of the common electrical charge of sub-pixels within a sampling area according to an embodiment of this disclosure.

[0063] Figure 8 is a schematic diagram of the common electrical charge of sub-pixels within a sampling area according to an embodiment of this disclosure.

[0064] Figure 9 is a schematic diagram of the common electrical charge of sub-pixels within a sampling area according to an embodiment of this disclosure.

[0065] Figure 10 is a framework diagram of the first initial model in the model training method of this disclosure embodiment.

[0066] Figure 11 is a flowchart of step S2 of the model training method according to an embodiment of the present disclosure.

[0067] Figure 12 is a flowchart of step S22 of the model training method according to an embodiment of the present disclosure.

[0068] Figure 13 is a flowchart of steps S3-S4 of the model training method according to an embodiment of the present disclosure.

[0069] Figure 14 is a framework diagram of the second initial model in the model training method of this disclosure embodiment.

[0070] Figure 15 is a flowchart of step S4 of the model training method according to an embodiment of the present disclosure.

[0071] Figure 16 is a schematic diagram of the model training device according to an embodiment of the present invention.

[0072] Figure 17 is a schematic diagram of the computer device provided in the embodiments of this disclosure. Detailed Implementation

[0073] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0074] Unless otherwise defined, the technical or scientific terms used in this disclosure shall have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms “first,” “second,” and similar terms used in this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an,” “a,” or “the,” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “including,” “comprising,” or “containing,” and similar terms mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. The terms “connected,” “linked,” or similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms “upper,” “lower,” “left,” and “right,” etc., are used only to indicate relative positional relationships, and these relative positional relationships may change accordingly when the absolute position of the described objects changes.

[0075] Figure 2 is a schematic diagram of an exemplary display panel; as shown in Figure 2, the display panel includes pixel units arranged in an array, each pixel unit including three colors of sub-pixels, the three colors being a first color, a second color, and a third color; in this embodiment, taking blue as the first color, green as the second color, and red as the third color as an example, the corresponding first color sub-pixel is called blue sub-pixel B, the second color sub-pixel is called green sub-pixel G, and the third color sub-pixel is called red pixel.

[0076] The inventors discovered that as refresh rates continue to increase, to ensure sufficient charging time for each sub-pixel, sub-pixels need to share power, meaning the pixel driving circuits of two sub-pixels will be electrically connected. As shown in Figure 3, the two blue sub-pixels B highlighted in Figure 3 are called the first blue sub-pixel B1 and the second blue sub-pixel B2. The sub-pixel connected to the first blue sub-pixel B1 is a sub-pixel of the same color, called the third blue sub-pixel B3. The pixel connected to the second blue sub-pixel B2 is the red sub-pixel R. When the same grayscale value is written to each blue sub-pixel B in the display panel, but if the grayscale value written to the red sub-pixel R connected to the second blue sub-pixel B2 is different from the grayscale value of the second blue sub-pixel B2, this will inevitably lead to a difference in brightness between the first blue sub-pixel B1 and the second blue sub-pixel B2, resulting in some textures on the display panel, such as horizontal or vertical stripes. To address this issue, a compensation table is typically created through manual testing. However, this method is time-consuming and labor-intensive, and the horizontal lines on different screens within the same batch are not identical. Therefore, creating a compensation table for each screen to obtain good display quality is not feasible for industrial screen production and is not easy to implement.

[0077] To address the aforementioned technical problems, this disclosure provides a model training method. The brightness compensation model trained using this method can generate corresponding brightness compensation tables for each color channel of any display panel. The color channels include a red channel, a green channel, and a blue channel. The red channel provides a grayscale voltage to the red sub-pixel R, enabling the red sub-pixel R to display the corresponding grayscale brightness data channel. Similarly, the green channel provides a grayscale voltage to the green sub-pixel G, enabling the green sub-pixel G to display the corresponding grayscale brightness data channel. The blue channel provides a grayscale voltage to the blue sub-pixel B, enabling the blue sub-pixel B to display the corresponding grayscale brightness data channel.

[0078] Figure 4 is a flowchart of the model training method according to an embodiment of the present disclosure; as shown in Figure 4, the model training method according to an embodiment of the present disclosure specifically includes the following steps:

[0079] S1. Obtain the first training sample set; the first training sample set includes multiple first sample groups; the first sample group includes image information of standard sub-pixels and image information of sub-pixels to be compensated; the standard sub-pixel is the sub-pixel that shares the same grayscale value as the sub-pixel with the same electrical charge; the sub-pixel to be compensated is the sub-pixel that shares the same electrical charge as the sub-pixel with a different color.

[0080] Taking Figure 2 as an example again, the grayscale value written to each blue sub-pixel B is 32, and the grayscale value written to the red sub-pixel is 0. Since the first blue sub-pixel B1 and the third blue sub-pixel B3 are connected, and the second blue sub-pixel B2 and the red sub-pixel R are connected, the first blue sub-pixel B1 is used as the standard sub-pixel, and the second blue sub-pixel B2 is used as the sub-pixel to be compensated. The image information displayed by the first blue sub-pixel B1 is the image information of the standard sub-pixel, and the image information displayed by the second blue sub-pixel B2 is the image information of the sub-pixel to be compensated.

[0081] S2. Use the first sample group as the input of the first initial model to be trained, and obtain the brightness compensation value of the sub-pixel to be compensated output by the first initial model; based on the brightness compensation value of the sub-pixel to be compensated and the preset standard brightness compensation value, train the first initial model to obtain the brightness compensation model.

[0082] In some examples, in step S2, the image information of the standard sub-pixel and the image information of the sub-pixel to be compensated in each sample group are input into the first initial model to be trained. The first initial model then outputs the brightness compensation value of the sub-pixel to be compensated. The brightness compensation value of the sub-pixel to be compensated and the preset standard brightness compensation value are input into the loss function to predict the loss. The parameters of the first initial model are updated through backpropagation. When the loss function becomes smaller and smaller until the first initial model converges, the training of the first initial model is completed, that is, the brightness compensation model is obtained.

[0083] The brightness compensation model obtained by the model training method in this embodiment analyzes the pixel driving circuit of each sub-pixel in the display panel, that is, analyzes the common electrical situation of each sub-pixel, samples a specific area of ​​the display panel, obtains sub-pixels that meet the conditions in the sampled area as standard sub-pixels and sub-pixels to be compensated, and extracts the image information of the standard sub-pixels and the image information of the sub-pixels to be compensated by controlling the grayscale values ​​written to each sub-pixel, and inputs them into the brightness compensation model to obtain the compensation table of the color channel of the sub-pixel to be compensated.

[0084] In some examples, Figure 5 is a detailed flowchart of step S1 of the model training method according to an embodiment of the present disclosure; as shown in Figure 5, step S1, obtaining the first training sample set, may specifically include:

[0085] S11. Based on the connection relationship of the driving circuits of each sub-pixel in the display panel, determine multiple sampling areas of the display panel, as well as the position information of the standard sub-pixel and the sub-pixel to be compensated in the sampling area, and store them.

[0086] It should be noted that the connection relationship of the driving circuit of each sub-pixel in the display panel is pre-designed. Therefore, the sampling area and the position information of the standard sub-pixel and the sub-pixel to be compensated that meet the conditions can be directly determined according to the connection relationship of the driving circuit of each sub-pixel. At this time, the sampling area can be marked and the position information of the standard sub-pixel and the sub-pixel to be compensated in the sampling area can be stored.

[0087] In some examples, the sub-pixels selected in the sampling region of this disclosure can be 4 rows and 6 columns of sub-pixels. For the position information of the standard sub-pixels and the position information of the sub-pixels to be compensated, a coordinate system can be established using two straight lines that run through the center of the sampling region and are perpendicular to each other as coordinate axes. In this case, the 4 rows and 6 columns of sub-pixels are evenly distributed in four fields. As shown in Figure 6, the position information of the standard sub-pixels and the position information of the sub-pixels to be compensated are (3, 2) and (1, -2), respectively.

[0088] S12. Control the display panel to display according to the preset lighting mode, obtain the display image of the sampling area in the display panel, and extract the image information of the standard sub-pixel and the image information of the sub-pixel to be compensated according to the stored position information of the standard sub-pixel and the sub-pixel to be compensated to obtain the first sample group. Multiple first sample groups constitute the first training sample set.

[0089] In some examples, to obtain the brightness compensation table for each color channel, it is necessary to control each sub-pixel in the display panel to display a specific grayscale value in order to obtain the first sample group. In step S22, after controlling the display panel to display according to a preset lighting mode, the display image in the marked sampling area of ​​the display panel can be acquired by an image acquisition module such as a camera. Then, based on the pre-stored position information of the standard sub-pixels and the sub-pixels to be compensated in the sampling area, the image information of the standard sub-pixels and the image information of the sub-pixels to be compensated are extracted to obtain the first sample group. It is understood that the function of the model obtained by model training depends on the training samples. Therefore, the process of acquiring the first training samples in this embodiment of the present disclosure is described in detail.

[0090] Each of the three color channels of the display panel has a brightness compensation table, as shown in the table below. The table below shows the compensation table for the blue channel. For example, the first and second preset grayscale values ​​can both be [0, 32, 64, 96, 127, 192, 224, 255]. The following explains the first sample set required for the brightness compensation model to output the compensation tables for the blue, green, and red channels.

[0091] First, the sample group of the compensation table for the blue channel that enables the brightness compensation model to output is explained. Taking the sampling area shown in Figure 3 as an example, the generation order of the compensation table is to first obtain the diagonal compensation values, and then generate the row and column compensation values. It should be noted that, due to the obvious pure blue horizontal stripe phenomenon, and because the compensated brightness must conform to the brightness rules, the grayscale values ​​of the red sub-pixel R, green sub-pixel G, and blue sub-pixel B are set to be equal, with R = B = G = [0, 32, 64, 96, 127, 192, 224, 255] as the standard brightness value. That is, the compensation value on the diagonal of the blue channel compensation table is 0, as shown in the table above.

[0092] Referring to Figure 3, the second blue sub-pixel B2 is the sub-pixel to be compensated, and the first blue sub-pixel B1 is the standard sub-pixel. At this point, the grayscale value of the blue sub-pixel B is controlled to be 32. The grayscale values ​​of the red sub-pixel R and the blue sub-pixel B are equal, and the values ​​are taken sequentially as [0, 32, 64, 96, 127, 192, 224, 255]. This yields the compensation value when the blue channel grayscale value is 32. Similarly, the grayscale value of the blue sub-pixel B is 64. The grayscale values ​​of the red sub-pixel R and the blue sub-pixel B are equal, and the values ​​are taken sequentially as [0, 32, 64, 96, 127, 192, 224, 255]. This yields the compensation value when the blue channel grayscale value is 64. Thus, the brightness compensation table for the blue channel can be obtained.

[0093] Of course, the grayscale values ​​of the red sub-pixel R and the blue sub-pixel B can also be controlled to 64, and the grayscale values ​​of the blue sub-pixel B can be taken as [0, 32, 64, 96, 127, 192, 224, 255] in sequence. At this time, the compensation value when the grayscale value of the blue channel is 64 can be obtained. Similarly, the brightness compensation table of the blue channel can also be obtained.

[0094] Next, we will explain the sample set of compensation tables that enable the brightness compensation model to output the green channel.

[0095] Figure 7 is a schematic diagram of the common electrical charge of sub-pixels within a sampling area according to an embodiment of this disclosure. As shown in Figure 7, the blue sub-pixel B framed in the sampling area is referred to as the first blue sub-pixel B1, and the sub-pixels connected to the first blue sub-pixel are the second blue sub-pixel B2 and the green sub-pixel G. The first blue sub-pixel B1 is used as the standard sub-pixel, and the green sub-pixel G is the sub-pixel to be compensated. The grayscale value of each red sub-pixel R in the control display panel is 0, and the green sub-pixel G and the blue sub-pixel B display the same grayscale value, that is, B = G = [0, 32, 64, 96, 127, 192, 224, 255]. At this time, the brightness compensation value of the diagonal of the brightness compensation table of the green sub-pixel G can be obtained.

[0096] Figure 8 is a schematic diagram of the common electrical charge of sub-pixels in a sampling area according to an embodiment of this disclosure. As shown in Figure 8, the two green sub-pixels G enclosed in the sampling area are referred to as the first green sub-pixel G1 and the second green sub-pixel G2. The first green sub-pixel G1 and the first blue sub-pixel B1 share the same electrical charge. The first green sub-pixel G1 serves as the standard green sub-pixel G, and the second green sub-pixel G2 shares the same electrical charge with the red sub-pixel R. The green sub-pixel G and the blue sub-pixel B are controlled to write the same grayscale value G = B = 32. The grayscale value of the red sub-pixel R is selected sequentially as [0, 64, 127, 192, 224, 255]. Similarly, the brightness compensation values ​​of the brightness compensation table of the green sub-pixel G, excluding the diagonal values, can be obtained. These values, together with the brightness compensation values ​​of the diagonal values ​​of the brightness compensation table of the green sub-pixel G obtained above, constitute the brightness compensation table of the green sub-pixel G.

[0097] Figure 9 illustrates a common-electric connection method for sub-pixels in a sampling area according to an embodiment of this disclosure. As shown in Figure 9, the two green sub-pixels G enclosed in the sampling area are referred to as the first red sub-pixel R1 and the second red sub-pixel R2. The red sub-pixel R shared by the first red sub-pixel R1 is the third red sub-pixel R, and the sub-pixel shared by the second red sub-pixel R2 is the blue sub-pixel B. The same grayscale value is written to each red sub-pixel R, and the same grayscale value is written to the green sub-pixel G and the blue sub-pixel B. Specifically, the grayscale value written to the red sub-pixel R is 32. The green sub-pixel G and the blue sub-pixel B display the same grayscale value, i.e., B = G = [0, 32, 64, 96, 127, 192, 224, 255]. At this point, the brightness compensation value when the red channel grayscale value is 32 can be obtained. Similarly, by changing the grayscale value written to the red sub-pixel R, brightness compensation values ​​for other grayscale values ​​can be obtained, thereby obtaining the brightness compensation table for the red channel.

[0098] In some examples, the ratio of the image width to the subpixel width of the standard subpixel's image information is 0.6–0.8, for example, 0.7; the ratio of the image length to the subpixel length of the standard subpixel's image information is 0.6–0.8, for example, 0.7; the ratio of the image width to the subpixel width of the image information of the subpixel to be compensated is 0.6–0.8, for example, 0.7; and the ratio of the image length to the subpixel length of the image information of the subpixel to be compensated is 0.6–0.8, for example, 0.7. The image information of the standard subpixel and the subpixel to be compensated are selected in this way to avoid visual interference from the subpixel edge brightness.

[0099] In some examples, Figure 10 is a framework diagram of the first initial model in the model training method of this disclosure embodiment; as shown in Figure 10, the first initial model includes: a first feature extraction network CNN1, a second feature extraction network CNN2, and an output network. Figure 11 is a detailed flowchart of step S2 of the model training method of this disclosure embodiment; as shown in Figure 11, step S2, which uses the sample group as input to the first initial model to be trained and obtains the brightness compensation value of the sub-pixel to be compensated output by the first initial model, includes:

[0100] S21. Use the image information of the standard sub-pixels in the sample group as the first feature extraction network CNN1 to obtain the image features of the standard sub-pixels output by the prime number first feature extraction network CNN1; use the image information of the sub-pixels to be compensated in the sample group as the second feature extraction network CNN2 to obtain the image features of the sub-pixels to be compensated output by the prime number second feature extraction network CNN2.

[0101] S22. Input the image features of the standard sub-pixel and the image features of the sub-pixel to be compensated into the output network, and regress the brightness compensation value of the sub-pixel to be compensated.

[0102] In some examples, Figure 12 is a detailed flowchart of step S22 of the model training method according to an embodiment of the present disclosure; as shown in Figure 12, the output network in step S22 includes a splicing layer cat, a first fully connected layer fc1, and a second fully connected layer fc2. Specifically, step S22 may include:

[0103] S221. Input the image features of the standard sub-pixels and the image features of the sub-pixels to be compensated into the stitching layer cat for feature stitching to obtain stitched features.

[0104] S222. Input the spliced ​​features into the first fully connected layer fc1 for feature operation to obtain the first intermediate features.

[0105] S223. Input the first intermediate feature into the second fully connected layer fc2 for feature operation and regress the brightness compensation value of the sub-pixel to be compensated.

[0106] In some examples, Figure 13 is a flowchart of steps S3-S4 of the model training method according to an embodiment of the present disclosure; as shown in Figure 13, in step S22 above, the display panel is controlled to display according to a preset lighting mode, and while acquiring the display image of the sampling area in the display panel, the method also includes acquiring the stripe pattern of the display image of the sampling area; the model training method further includes:

[0107] S3. Generate a second sample group based on the displayed image of the sampling area and the corresponding stripe pattern. Multiple second sample groups constitute the second training sample set.

[0108] In some examples, step S3 specifically includes performing at least data processing on the display image of the sampled area, such as data enhancement through rotation and chromaticity transformation, to obtain a processed display image, and using the processed display image and the corresponding stripe pattern as a second sample group.

[0109] Specifically, if the displayed image of the sampled area has texture, it can be labeled as 1; if the displayed image of the sampled area has no texture, it can be labeled as 0.

[0110] S4. Use the displayed image in the second sample group as the input of the second initial model to obtain the predicted stripe pattern of the second initial model output. Based on the predicted stripe pattern and the stripe pattern in the second sample group, train the second model to obtain the classification model.

[0111] In some examples, Figure 14 is a framework diagram of the second initial model in the model training method of this disclosure embodiment; as shown in Figure 14, the second initial model includes a third feature extraction network CNN3, a fourth feature extraction network CNN4, a fifth feature extraction network CNN5, a third fully connected layer fc3, a fourth fully connected layer fc4, and a classification layer connected in sequence. Figure 15 is a detailed flowchart of step S4 of the model training method of this disclosure embodiment; as shown in Figure 15, step S4 may specifically include the following steps:

[0112] S41. Input the displayed image of the sampling area into the third feature extraction network CNN3, and perform feature extraction through the third feature extraction network CNN3, the fourth feature extraction network CNN4, and the fifth feature extraction network CNN5.

[0113] Specifically, the third feature extraction network (CNN3), the fourth feature extraction network (CNN4), and the fifth feature extraction network (CNN5) can be convolutional neural networks. The kernel size, number of kernels, and stride of these three convolutional neural networks are as follows: CNN3 has a kernel size of 3, a kernel count of 3, and a stride of 2; CNN4 has a kernel size of 3, a kernel count of 128, and a stride of 2; and CNN5 has a kernel size of 3, a kernel count of 256, and a stride of [missing information].

[0114] S42. The output features of the fifth feature extraction network CNN5 are processed through the third fully connected layer fc3 to obtain the second intermediate features.

[0115] S43. Perform feature operations on the second intermediate feature through the fourth fully connected layer fc4 to obtain the third intermediate feature.

[0116] S44. The classification layer predicts the stripe pattern of the displayed image of the sampling area based on the third intermediate feature input, and obtains the predicted stripe pattern. Based on the predicted stripe pattern and the stripe pattern in the second sample group, the second model is trained to obtain the classification model.

[0117] In some examples, the classification layer can specifically be a softmax (normalized exponential function). Stripes refer to whether the displayed image has stripes or not; therefore, the predicted stripe situation output by the classification layer is either 1 (representing stripes) or 0 (representing no stripes). In step S44, the stripe situations in the second sample group and the predicted stripes are used to calculate the loss using MSE loss, and the model parameters of the second initial model are updated through backpropagation until the classification model is obtained.

[0118] The classification model trained using the above method is approximately 271MB in size and can process images of 2048*2048 pixels (the model can be even smaller if the input image is reduced). Processing speed is hardware-dependent; on a typical machine with a GPU, it can generally achieve millisecond-level processing speed. The classification accuracy of the model can reach 95%.

[0119] The model training method of this disclosure can obtain a brightness compensation model and a classification model. Therefore, by lighting up the display panel and then sampling according to a preset sampling area, a brightness compensation table for the three color channels of the display panel can be obtained using the brightness compensation model. Next, the brightness of the displayed image on the display panel can be compensated using the brightness compensation table, thus obtaining a compensated image. Finally, image acquisition can be performed on the compensated image according to the sampling area, and the displayed image of the sampling area can be used as input to the classification model. Based on the stripe pattern of the displayed image of the sampling area output by the classification model, the brightness compensation effect of the brightness compensation table generated by the brightness compensation model can be verified.

[0120] Accordingly, this disclosure also provides a display compensation method, which includes generating a brightness compensation table for each color channel of a display panel using a brightness compensation model; the brightness compensation model is trained by the model training method described above; compensating the display data of sub-pixels in the display panel based on the brightness compensation table; and controlling each sub-pixel in the display panel to display based on the compensated display data.

[0121] The display data in this embodiment refers to the grayscale voltage used for sub-pixel display. The color channels include a red channel, a green channel, and a blue channel. The red channel provides a grayscale voltage to the red sub-pixel R, enabling the red sub-pixel R to display the corresponding grayscale brightness data channel. Similarly, the green channel provides a grayscale voltage to the green sub-pixel G, enabling the green sub-pixel G to display the corresponding grayscale brightness data channel. The blue channel provides a grayscale voltage to the blue sub-pixel B, enabling the blue sub-pixel B to display the corresponding grayscale brightness data channel.

[0122] Because the display compensation method in this embodiment uses the brightness compensation model trained by the above method to generate a brightness compensation table to compensate the display data, the display screen can effectively avoid the problem of stripes.

[0123] Accordingly, this disclosure also provides a display panel including a plurality of sub-pixels, a driving module, and a storage module storing a brightness compensation table. The plurality of sub-pixels can be sub-pixels to which analog signals (e.g., data voltages) generated by processing digital image signals are applied. Each sub-pixel emits light according to the applied voltage (e.g., as in an OLED display), or displays an image by controlling the intensity of light passing through the liquid crystal according to the liquid crystal transmittance (e.g., as in an LCD display), wherein the liquid crystal transmittance is adjustable by the applied voltage. For example, an OLED display panel mainly includes a pixel driving circuit and a light-emitting device, the pixel driving circuit driving the light-emitting device to emit light, thereby realizing image display.

[0124] In some examples, the brightness compensation table in this disclosure embodiment can be a LUT table, which can be a two-dimensional table or a three-dimensional table, and the LUT table stores the brightness compensation value of each color channel at different gray levels.

[0125] Figure 16 is a schematic diagram of the model training device according to an embodiment of the present invention. As shown in Figure 16, this embodiment of the present invention also provides a model training device, which includes a sample acquisition module 10 and a training module 20. The sample acquisition module can be used to execute the above-described step S1, and the training module can be used to execute the above-described step S2. Specifically, the sample acquisition module 10 is configured to acquire a first training sample set; the first training sample set includes multiple first sample groups; each first sample group includes image information of a standard sub-pixel and image information of a sub-pixel to be compensated; the standard sub-pixel is a sub-pixel that shares the same grayscale value as a sub-pixel with the same electrical charge; the sub-pixel to be compensated is a sub-pixel that shares the same electrical charge as a sub-pixel with a different color; the training module 20 is configured to use the sample group as input to a first initial model to be trained, obtain the brightness compensation value of the sub-pixel to be compensated output by the first initial model, and train the first initial model based on the brightness compensation value of the sub-pixel to be compensated and a preset standard brightness compensation value to obtain a brightness compensation model.

[0126] Figure 17 is a schematic diagram of the computer device provided in the embodiments of this disclosure. As shown in Figure 17, the computer device 100 may be a terminal, such as a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness device, personal digital assistant, etc.; or it may be a server.

[0127] Referring to FIG13, computer device 100 may include one or more of the following components: processing component 102, memory 104, power supply component 106, multimedia component 108, audio component 110, input / output (I / O) interface 112, and communication component 116.

[0128] Processing component 102 typically controls the overall operation of computer device 100, such as operations associated with display, telephone calls, data communication, camera operation, and recording operations. Processing component 102 may include one or more processors 120 to execute instructions to perform all or part of the steps of the methods described above. Furthermore, processing component 102 may include one or more modules to facilitate interaction between processing component 102 and other components. For example, processing component 102 may include a multimedia module to facilitate interaction between multimedia component 108 and processing component 102.

[0129] Memory 104 is configured to store various types of data to support the operation of computer device 100. Examples of such data include instructions for any application or method operating on computer device 100, contact data, phone book data, messages, pictures, videos, etc. Memory 104 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0130] Power supply component 106 provides power to various components of computer device 100. Power supply component 106 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to computer device 100.

[0131] Multimedia component 108 includes a screen that provides an output interface between the computer device 100 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 108 includes a front-facing camera and / or a rear-facing camera. When the computer device 100 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0132] Audio component 110 is configured to output and / or input audio signals. For example, audio component 110 includes a microphone (MIC) configured to receive external audio signals when computer device 100 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 104 or transmitted via communication component 116. In some embodiments, audio component 110 also includes a speaker for outputting audio signals.

[0133] I / O interface 112 provides an interface between processing component 102 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0134] Communication component 116 is configured to facilitate wired or wireless communication between computer device 100 and other devices. Computer device 100 can access wireless networks based on communication standards, such as WiFi, 2G or 3G, 4G LTE, 5G NR, or combinations thereof. In one exemplary embodiment, communication component 116 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 116 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0135] In an exemplary embodiment, the computer device 100 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0136] This disclosure also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the instruction recognition method or natural language model training method described in the above embodiments. For example, the computer-readable storage medium may include the storage component of a tablet computer, a hard disk of a personal computer, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), optical disc read-only memory (CD-ROM), flash memory, or any combination of the above storage media, or other suitable storage media.

[0137] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

[0138] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0139] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A model training method comprising: obtaining a first training sample set; the first training sample set comprises a plurality of first sample groups; the first sample group comprises image information of a standard sub-pixel and image information of a to-be-compensated sub-pixel; the standard sub-pixel is a sub-pixel whose written gray scale value is the same as that of a sub-pixel of a different color; the to-be-compensated sub-pixel is a sub-pixel that is common to a sub-pixel of a different color; taking the first sample group as input of a first initial model to be trained, to obtain a brightness compensation value of the to-be-compensated sub-pixel output by the first initial model; training the first initial model based on the brightness compensation value of the to-be-compensated sub-pixel and a preset standard brightness compensation value, to obtain a brightness compensation model. 2.The model training method of claim 1, wherein, the first initial model comprises a first feature extraction network, a second feature extraction network, and an output network; the step of taking the sample group as input of the first initial model to be trained, to obtain a brightness compensation value of the to-be-compensated sub-pixel output by the first initial model, comprises: taking the image information of the standard sub-pixel in the sample group as input of the first feature extraction network, to obtain image features of the standard sub-pixel output by the first feature extraction network; taking the image information of the to-be-compensated sub-pixel in the sample group as input of the second feature extraction network, to obtain image features of the to-be-compensated sub-pixel output by the second feature extraction network; inputting the image features of the standard sub-pixel and the image features of the to-be-compensated sub-pixel into the output network, to regress the brightness compensation value of the to-be-compensated sub-pixel. 3.The model training method of claim 2, wherein, the output network comprises a concatenation layer, a first full connection layer, and a second full connection layer; the step of inputting the image features of the standard sub-pixel and the image features of the to-be-compensation sub-pixel into the output network, to regress the brightness compensation value of the to-be-compensation sub-pixel, comprises: inputting the image features of the standard sub-pixel and the image features of the to-be- compensation sub-pixel into the concatenation layer for feature concatenation, to obtain concatenated features; inputting the concatenated features into the first full connection layer for feature operation, to obtain first intermediate features; inputting the first intermediate features into the second full connection layer for feature operation, to regress the brightness compensation value of the to-be-compensation sub-pixel. 4.The model training method of claim 1, wherein, the step of obtaining the first training sample set comprises: determining a plurality of sampling regions of the display panel and position information of the standard sub-pixels and the to-be-compensated sub-pixels in the sampling regions according to a connection relationship of driving circuits of the sub-pixels in the display panel, and storing the same; controlling the display panel to display in a preset lighting mode, obtaining display images of the sampling regions in the display panel, and extracting image information of the standard sub-pixels and image information of the to-be-compensated sub-pixels according to the stored position information of the standard sub-pixels and the to-be-compensated sub-pixels, to obtain the first sample group, and a plurality of the first sample groups constitute the first training sample set. 5.The model training method of claim 4, wherein, the sub-pixels of the display panel comprise sub-pixels of a first color, a second color, and a third color; The standard sub-pixel and the to-be-compensated sub-pixel in part of the sample groups satisfy: the standard sub-pixel and the sub-pixel sharing the same color, the color of the standard sub-pixel and the to-be-compensated sub-pixel is the first color; the to-be-compensated sub-pixel shares the same color with the third color sub-pixel. The standard sub-pixel and the to-be-compensated sub-pixel in part of the samples groups satisfy: the standard sub-pixel and the sub-pixel sharing the same color, the standard sub-pixel and the to-be-compensated sub-pixel are the same color; the to-be-compensated sub-pixel shares the same color with the third color of the sub-pixel. The standard sub-pixel and the to-be-compensated sub-pixel of part of the sample groups satisfy: the standard sub-pixel is the first color, and the to-be-compensated sub-pixel is the second color; the to-be-compensated sub-pixel and the standard sub-pixel share the same color. The standard sub-pixel and the to-be-compensated sub-pixel in part of sample groups satisfy: the standard sub-pixel and the sub-pixel sharing the same color, and the color of the standard sub-pixel and the to-be-compensated sub-pixel is third color; the to-be-compensated sub-pixel shares the same color with the first color of the sub-pixel. 6.The model training method of claim 5, wherein, When the standard sub-pixel and the to-be-compensated sub-pixel in the sample group satisfy: the standard sub-pixel and the sub-pixel sharing the same color, the color standard sub-pixel and the to-be-compensated sub-pixel is the first color; The to-be-compensated sub-pixel shares the same color with the third color of the subpixel, the control display panel displays according to the preset lighting mode, obtains the display image of the sampling area in the display panel, and extracts the image information of the standard sub-pixel and the image information of the to-be-compensated sub-pixel according to the stored position information of the standard sub-pixel and the to-be-compensated sub-pixel, to obtain the first sample group. The first color sub-pixel in the display panel is controlled to write the same first preset gray scale value, the second color sub-pixel and the third color sub-pixel are controlled to write the same second preset gray scale value, the display image of the sampling area in the display panel is obtained, and the image information of the standard sub-pixel and the image information of the to-be-compensate sub-pixel are extracted according to the stored position information of the standard sub-pixel and the to-be-compensated pixel, to obtain the first sample group. 7.The method of Claim 5, wherein, When the standard sub-pixel and the to-be-compensated sub-pixel in the first sample group satisfy: the standard sub-pixel is the first color, and the to-be-compensated pixel is the second color; the to-be-compensated sub-pixel and the standard sub-pixels share the same color, the control display panel displays according to the preset lighting mode, obtains the display image of the sample area in the display panel, and extracts the image information of the standard sub-pixel and the to-be-compensated sub-pixel according to the stored position information of the standard subpixel and the to-be-compensated sub-pixel, to obtain the first sample group. controlling a third color sub-pixel in the display panel to write 0 gray scale, controlling the first color sub-pixel and the second color sub-pixel to write the same preset gray scale value, obtaining a display image of the sampling area in the display panel, and extracting image information of the standard sub-pixel and image information of the to-be-compensated sub-pixel according to the stored position information of the standard sub-pixel and the to-be-compensated sub-pixel, to obtain the first sample group. 8.The method of Claim 5, wherein, The standard sub-pixel and the to-be-compensated sub-pixel in the sample group satisfy: the standard sub-pixel and a sub-pixel sharing the same electrode are of different colors, and the standard sub-pixel and the to-be-compensated sub-pixel are of the same gray scale value; the standard sub-pixel is of the second color, and the sub-pixel sharing the same electrode with the standard sub-pixel is of the first color; the to-be-compensated sub-pixel is of the third color; when the to-be-compensated sub-pixel shares the same electrode with the second color sub-pixel, the display panel is controlled to display according to a preset lighting mode, a display image of the sampling area in the display panel is obtained, and image information of the standard sub-pixel and image information of the to-be-compensated sub-pixel are extracted according to the stored position information of the standard sub-pixel and the to-be-compensated sub-pixel, to obtain the first sample group. controlling a third color sub-pixel in the display panel to write 0 gray scale, controlling the first color sub-pixel and the second color sub-pixel to write the same preset gray scale value, obtaining a display image of the sampling area in the display panel, and extracting image information of the standard sub-pixel and image information of the to-be-compensated sub-pixel according to the stored position information of the standard sub-pixel and the to-be-compensated sub-pixel, to obtain the first sample group. 9.The method of Claim 5, wherein, When the standard sub-pixel and the to-be-compensated sub-pixel in the sample group satisfy: the standard sub-pixel and a sub-pixel sharing the same electrode are of the same color, and the standard sub-pixel and the to-be-compensated sub-pixel are of the third color; when the to-be-compensated sub-pixel shares the same electrode with the first color sub-pixel, the display panel is controlled to display according to a preset lighting mode, a display image of the sampling area in the display panel is obtained, and image information of the standard sub-pixel and image information of the to-be-compensated sub-pixel are extracted according to the stored position information of the standard sub-pixel and the to-be-compensated sub- pixel, to obtain the first sample group. controlling a third color sub-pixel in the display pane to display the same first preset gray scale value, controlling the first color sub-pixel to display the same second preset gray scale value, obtaining a display image of the sampling area in the display panel, and according to the stored position information of the standard sub-pixel and the to-be-compensated, extracting image information of the standard sub-pixel and image information of the to-be-compensated, to obtain the first sample group. 10.The model training method of any one of claims 1-9, wherein, The image width of the image information of the standard sub-pixel and the width of the sub-pixel are in a ratio of 0.6-0.8; the image length of the image information of the standard sub-pixel and the length of the sub-pixel are in a ratio of 0.6-0.8; the image width of the image information of the sub-pixel to be compensated and the width of the sub-pixel are in a ratio of 0.6-0.8; and the image length of the image information of the sub-pixel to be compensated and the length of the sub-pixel are in a ratio of 0.6-0.

8. 11.The method of Claim 4, wherein, The display panel is controlled to display in a preset lighting mode, and the display image of the sampling area in the display panel is obtained, and the display image of the sampling area also includes a stripe condition; the model training method further includes: generating a second sample group based on the display image of the sampling area and the corresponding stripe condition, and multiple second sample groups constitute a second training sample set; taking the display image in the second sample group as the input of a second initial model to obtain a predicted stripe condition output by the second initial model, and training the second model according to the predicted stripe condition and the stripe condition in the second sample group to obtain a classification model. 12.The model training method of claim 11, wherein, The second initial model includes a third feature extraction network, a fourth feature extraction network, a fifth feature extraction network, a third full connection layer, a fourth full connection layer and a classification layer connected in sequence; and the step of taking the display image in the second sample group as the input of the second initial model to obtain a predicted stripe condition output by the second initial model, and training the second model according to the predicted stripe condition and the stripe condition in the second sample group to obtain a classification model includes: inputting the display image of the sampling area into the third feature extraction network, and performing feature extraction on the display image via the third feature extraction network, the fourth feature extraction network and the fifth feature extraction network; performing feature operation on the output feature of the fifth feature extraction network via the third full connection layer to obtain a second intermediate feature; performing feature operation on the second intermediate feature via the fourth full connection layer to obtain a third intermediate feature; the classification layer inputs the third intermediate feature to predict the stripe condition of the display image of the sampling area to obtain a predicted stripe condition, and the second model is trained according to the predicted stripe condition and the stripe condition in the second sample group to obtain a classification model. 13.The method of Claim 11, wherein, The step of generating a second sample group based on the display image of the sampling area and the corresponding stripe condition includes: performing data processing on the display image of the sampling area to obtain a processed display image, and taking the processed display image and the corresponding stripe condition as a second sample group. 14.The method of Claim 1, wherein, The step of training the first initial model based on the brightness compensation value of the sub-pixel to be compensated and a preset standard brightness compensation value to obtain a brightness compensation model includes: The first initial model is trained based on the luminance compensation value of the to-be-compensated sub-pixel and a preset standard luminance compensation value, and a loss function is used to adjust the model parameters of the first initial model according to an output result of the loss function until the loss function converges, so as to obtain the luminance compensation model.

15. A display compensation method, comprising: generating a luminance compensation table of each color channel of a display panel by using a luminance compensation model; the luminance compensation model is obtained by using the model training method in any one of claims 1-14; compensating display data of sub-pixels in the display panel based on the luminance compensation table; controlling each sub-pixel in the display panel to display based on the compensated display data.

16. A model training apparatus, comprising: a sample acquisition module configured to acquire a first training sample set; the first training sample set comprises a plurality of first sample groups; the first sample group comprises image information of a standard sub-pixel and image information of a to-be-compensated sub-pixel; the standard sub-pixel is a sub-pixel written with the same gray scale value as a sub-pixel with a different color; the to-be-compensated sub-pixel is a sub-pixel with the same color as a sub-pixel; a training module configured to take the sample group as an input of a first initial model to be trained, obtain a luminance compensation value of the to-be-compensated sub-pixel output by the first initial model, and train the first initial model based on the luminance compensation value of the to-be-compensated sub-pixel and a standard luminance compensation value to obtain a luminance compensation model.

17. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor to implement the method in any one of claims 1-14.

18. A computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method in any one of claims 1-14.

Images