Method, device, electronic device, and storage medium for automatically learning network preprocessing parameters

Abstract

The present application discloses a method, device, electronic device, and storage medium for automatically learning network preprocessing parameters, wherein the method includes training a semantic segmentation model through randomly constructed image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each picture, and the image processing parameter items include image processing parameter values; inputting multiple pictures into a preset parameter preprocessing network to obtain image processing target parameter values ​​corresponding to each picture; adjusting the pictures to be semantically segmented to obtain multiple target pictures according to the image processing target parameter values ​​corresponding to each picture; inputting the multiple target pictures into the pre-trained semantic segmentation model to perform image semantic segmentation. This application improves the semantic segmentation effect of the semantic segmentation model during prediction.

Description

Technical Field

[0001] The present application relates to the field of computer vision technology, and in particular to a method, device, electronic device, and storage medium for automatically learning network preprocessing parameters. Background Art

[0002] Semantic segmentation, the process of linking each pixel in an image to a class label, is crucial in self-driving car technology because it is important for the model to understand the context of the environment in which it operates.

[0003] In related technologies, when using a trained semantic segmentation model for prediction, the data to be input into the model is not transformed in terms of brightness, contrast, saturation, etc. corresponding to the training. Instead, the data is directly input into the network for prediction, resulting in poor segmentation effect of the semantic segmentation model. Summary of the Invention

[0004] The embodiments of the present application provide a method, device, electronic device, and storage medium for automatically learning network preprocessing parameters to improve the effect of semantic segmentation during prediction.

[0005] The embodiments of this application adopt the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a method for automatically learning network preprocessing parameters, which is used for a semantic segmentation model. The method includes: training a semantic segmentation model through randomly constructed image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each picture, and the image processing parameter items include image processing parameter values; inputting multiple pictures into a preset parameter preprocessing network to obtain image processing target parameter values ​​corresponding to each picture; adjusting the pictures to be semantically segmented according to the image processing target parameter values ​​corresponding to each picture to obtain multiple target pictures; inputting the multiple target pictures into the pre-trained semantic segmentation model to perform image semantic segmentation.

[0007] In the second aspect, an embodiment of the present application also provides a device for automatically learning network preprocessing parameters for a semantic segmentation model, the device comprising: a semantic model training module, for training a semantic segmentation model through randomly constructed image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each picture, and the image processing parameter items include image processing parameter values; a preprocessing network module, for inputting multiple pictures into a preset parameter preprocessing network to obtain an image processing target parameter value corresponding to each picture; an adjustment module, for adjusting the pictures to be semantically segmented according to the image processing target parameter value corresponding to each picture to obtain multiple target pictures; a semantic segmentation module, for inputting the multiple target pictures into the pre-trained semantic segmentation model to perform image semantic segmentation.

[0008] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor; and a memory arranged to store computer-executable instructions, wherein the executable instructions, when executed, enable the processor to perform the above method.

[0009] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores one or more programs. When the one or more programs are executed by an electronic device including multiple application programs, the electronic device executes the above method.

[0010] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:

[0011] First, randomly construct image processing parameter items and then train to obtain a semantic segmentation model. Then, multiple images are input into a preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image. Furthermore, based on the image processing target parameter value corresponding to each image, the images to be semantically segmented are adjusted to obtain multiple target images. Finally, the multiple target images are input into the pre-trained semantic segmentation model to perform image semantic segmentation. By constructing a preset parameter preprocessing network to obtain multiple target images, and then inputting them into the semantic segmentation model to perform image semantic segmentation, the effect of semantic segmentation can be greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0013] Figure 1 Schematic diagram of a flow chart of a method for automatically learning network preprocessing parameters in an embodiment of the present application;

[0014] Figure 2 Schematic diagram of the structure of a device for automatically learning network preprocessing parameters in an embodiment of the present application;

[0015] Figure 3 This is a schematic diagram of IOU comparison in a method for automatically learning network preprocessing parameters in an embodiment of the present application;

[0016] Figure 4 This is a structural diagram of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0017] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0018] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0019] Semantic segmentation networks in related technologies typically undergo data preprocessing during training. Data preprocessing parameters are typically set within a certain range. During training, a random number within this range is generated and the image data fed into the network is subjected to transformations such as brightness, contrast, and saturation. However, when using the trained model for prediction, the data is not subjected to the corresponding brightness, contrast, and saturation transformations used during training. Instead, the data is directly fed into the network for prediction.

[0020] The present application embodiment provides a method for automatically learning network preprocessing parameters, such as Figure 1 As shown, a flow chart of a method for automatically learning network preprocessing parameters according to an embodiment of the present application is provided, wherein the method comprises at least the following steps S110 to S140:

[0021] Step S110 , training a semantic segmentation model by randomly constructing image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each picture, and the image processing parameter items include image processing parameter values.

[0022] First, a semantic segmentation model model1 is established and trained. When training the semantic segmentation model model1, the model is trained by using randomly constructed image processing parameter items as training parameter items.

[0023] It is understandable that those skilled in the art can obtain a semantic segmentation model through training based on relevant technologies, and perform some brightness, contrast, saturation and other transformations on the image data to be input into the network during the training process.

[0024] Furthermore, the image processing parameter items are defined as being used to adjust the display effect of each picture. That is, the image processing parameter items can adjust the display effects such as brightness, contrast, and saturation of each picture. At the same time, it can be understood that the image processing parameter items include image processing parameter values. For each of the image processing parameter items, there is a corresponding image processing parameter value. However, these image processing parameter values ​​have preset intervals, which are only introduced in the training stage in the related art. However, this application is not only about the introduction in the training stage, but also about adjusting the pictures to be input into the semantic segmentation model by constructing a network.

[0025] Step S120: Input multiple images into a preset parameter preprocessing network to obtain an image processing target parameter value corresponding to each image.

[0026] The semantic segmentation model model1 in the above steps is frozen after training, even though its parameters do not participate in the backpropagation process. However, it is understood that other functions in the model can still be performed.

[0027] Then, we start over and input multiple images into the preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image. The constructed preset parameter preprocessing network is used to process multiple images and output the image processing target parameter value corresponding to each image.

[0028] It can be understood that the multiple images here and the images used to train the semantic segmentation model can be from the same batch of images.

[0029] Step S130 , adjusting the pictures to be semantically segmented according to the image processing target parameter value corresponding to each picture to obtain a plurality of target pictures.

[0030] Based on the image processing target parameter values ​​corresponding to each image output by the parameter preprocessing network, the image to be semantically segmented can be adjusted to obtain multiple target images. The adjustment methods include, but are not limited to, adjusting the image processing parameter values ​​such as brightness, contrast, and saturation.

[0031] It can be understood that the multiple target images here are images that need to be semantically segmented, that is, images that need to be predicted.

[0032] Step S140: Input the plurality of target images into the pre-trained semantic segmentation model to perform image semantic segmentation.

[0033] The multiple target images are input into the pre-trained semantic segmentation model to perform image semantic segmentation and obtain corresponding semantic segmentation results. The image processing target parameter values ​​output by the preset parameter preprocessing network are adjusted before being input into the semantic segmentation model to improve the semantic segmentation prediction effect.

[0034] Preferably, in one embodiment of the present application, the network structure of the preset parameter preprocessing network includes at least two fully connected layers. In specific implementation, relevant experiments have found that the network structure of the automatic learning preprocessing parameter network using two fully connected layers has the best segmentation effect.

[0035] In one embodiment of the present application, multiple images are input into a preset parameter preprocessing network to obtain an image processing target parameter value corresponding to each image, including: inputting multiple images into a preset parameter preprocessing network, processing them sequentially through a two-dimensional normalization layer, a first fully connected layer, a one-dimensional normalization layer, an activation layer, and a second fully connected layer to obtain an image processing target parameter value corresponding to each image.

[0036] In specific implementation, the network structure of the preset parameter preprocessing network is:

[0037] Image input Input, HxW RGB images,

[0038] Two-dimensional normalization layer BatchNormalization2d,

[0039] The first fully connected layer is Linear, In: 3xHxW, Out: 64,

[0040] One-dimensional normalization layer BatchNorm1d,

[0041] Activation layer LeakyReLU,

[0042] The second fully connected layer is Linear, In: 64, Out: 3.

[0043] It should be noted that, except for the preferred use of two fully connected layers, the network structure of the preset parameter preprocessing network in the embodiments of the present application is not specifically limited, and those skilled in the art can make a choice based on actual usage.

[0044] In addition, if Figure 3The figure below shows an IOU comparison diagram for a method for automatically learning network preprocessing parameters in an embodiment of the present application. Line a represents the convolutional learning preprocessing parameter network, line b represents the two-layer fully connected learning preprocessing parameter network, and line c represents the original pretrained model. As can be seen, using two fully connected layers for the preprocessing network with preset parameters can produce better segmentation results.

[0045] In one embodiment of the present application, after the semantic segmentation model is trained by randomly constructed image processing parameter items, it also includes: the semantic segmentation network part does not perform back propagation when the preset parameter preprocessing network trains the image processing target parameter value corresponding to each picture.

[0046] During specific implementation, a semantic segmentation network is trained by randomly constructed image processing parameter items, so that the semantic segmentation network does not perform back propagation until the multiple target images are input into the pre-trained semantic segmentation model, and then back propagation is started.

[0047] In one embodiment of the present application, it also includes: performing data normalization processing on the image processing parameter value so that the image processing parameter value is within a preset range interval; the preset range interval is the same as the range interval of the random value corresponding to the randomly constructed image processing parameter item in the semantic segmentation model.

[0048] In a specific implementation, the image processing parameter values ​​are A, B, and C as an example for detailed description, wherein A, B, and C correspond to the brightness adjustment parameter A, the contrast adjustment parameter B, and the saturation adjustment parameter C.

[0049] In some embodiments, the three parameters A, B, and C need to be normalized before use, and their values ​​are normalized to the range of 0 to 1. Then, A, B, and C are each multiplied by a number M and then added to a number N, so that the values ​​of A, B, and C can be limited to the interval [N, M+N].

[0050] It should be noted that the [N, M+N] interval obtained in the above manner needs to be consistent with the interval of random values ​​of random brightness, random contrast, and random saturation set when training the semantic segmentation model model1, that is, it needs to be adjusted within its interval.

[0051] In one embodiment of the present application, the step of inputting multiple images into a preset parameter preprocessing network to obtain an image processing target parameter value corresponding to each image includes: inputting multiple images into a preset parameter preprocessing network to obtain an image processing target brightness parameter value, a target contrast parameter value, and a target saturation parameter value corresponding to each image.

[0052] In specific implementations, multiple images are fed into a pre-processing network with preset parameters to obtain target brightness, contrast, and saturation parameter values ​​for each image. For each image, the pre-processing network outputs a target brightness, contrast, and saturation parameter value. These values ​​are then fed into the semantic segmentation model model1 for semantic segmentation.

[0053] An embodiment of the present application also provides a semantic segmentation method, which adopts the method of automatically learning network preprocessing parameters to obtain a preset parameter preprocessing network, and inputs the preset parameter preprocessing network into a preset semantic segmentation model, thereby optimizing the segmentation results of the preset semantic segmentation model.

[0054] A semantic segmentation model is trained by randomly constructed image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each picture, and the image processing parameter items include image processing parameter values; multiple pictures are input into a preset parameter preprocessing network to obtain image processing target parameter values ​​corresponding to each picture; according to the image processing target parameter values ​​corresponding to each picture, the pictures to be semantically segmented are adjusted to obtain multiple target pictures; the multiple target pictures are input into the pre-trained semantic segmentation model to perform image semantic segmentation.

[0055] In the above semantic segmentation method, before inputting the plurality of images into the preset parameter preprocessing network, the method further includes:

[0056] The preset parameter preprocessing network is obtained through machine learning training using multiple sets of image data, wherein each set of image data includes: a sample image and an image pixel type corresponding to the sample image.

[0057] In the above semantic segmentation method, after the semantic segmentation model is trained by randomly constructing the image processing parameter items, the method further includes:

[0058] The semantic segmentation model does not perform back propagation until the multiple target images are input into the pre-trained semantic segmentation model, and then back propagation is started.

[0059] The semantic segmentation method further includes: performing data normalization processing on the image processing parameter value so that the image processing parameter value is within a preset range;

[0060] The preset range interval is the same as the range interval of the random value corresponding to the randomly constructed image processing parameter item in the semantic segmentation model.

[0061] In the above-mentioned semantic segmentation method, the process of inputting multiple images into a preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image includes:

[0062] Multiple images are input into the preset parameter preprocessing network to obtain the image processing target brightness parameter value, target contrast parameter value, and target saturation parameter value corresponding to each image.

[0063] In the above-mentioned semantic segmentation method, the network structure of the preset parameter preprocessing network includes at least two fully connected layers.

[0064] In the above semantic segmentation method, multiple images are input into the preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image, including:

[0065] Multiple images are input into the preset parameter preprocessing network, and processed in sequence through the two-dimensional normalization layer, the first fully connected layer, the one-dimensional normalization layer, the activation layer, and the second fully connected layer to obtain the image processing target parameter value corresponding to each image.

[0066] The embodiment of the present application also provides a device 200 for automatically learning network preprocessing parameters, such as Figure 2 As shown, a schematic diagram of the structure of an apparatus for automatically learning network preprocessing parameters in an embodiment of the present application is provided. The apparatus 200 for automatically learning network preprocessing parameters includes at least: a semantic model training module 210, a preprocessing network module 220, an adjustment module 230, and a semantic segmentation module 240, wherein:

[0067] In one embodiment of the present application, the semantic model training module 210 is specifically used to: first establish and train a semantic segmentation model model1, and when training the semantic segmentation model model1, train the model by using randomly constructed image processing parameter items as training parameter items.

[0068] It is understandable that those skilled in the art can obtain a semantic segmentation model through training based on relevant technologies, and perform some brightness, contrast, saturation and other transformations on the image data to be input into the network during the training process.

[0069] Furthermore, the image processing parameter items are defined as being used to adjust the display effect of each picture. That is, the image processing parameter items can adjust the display effects such as brightness, contrast, and saturation of each picture. At the same time, it can be understood that the image processing parameter items include image processing parameter values. For each of the image processing parameter items, there is a corresponding image processing parameter value. However, these image processing parameter values ​​have preset intervals, which are only introduced in the training stage in the related art. However, this application is not only about the introduction in the training stage, but also about adjusting the pictures to be input into the semantic segmentation model by constructing a network.

[0070] In one embodiment of the present application, the pre-processing network module 220 is specifically used to freeze the semantic segmentation model model1 after training, i.e., its parameters do not participate in the back-propagation process. However, it is understood that other functions in the model can be performed.

[0071] Then, we start over and input multiple images into the preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image. The constructed preset parameter preprocessing network is used to process multiple images and output the image processing target parameter value corresponding to each image.

[0072] It can be understood that the multiple images here and the images used to train the semantic segmentation model can be from the same batch of images.

[0073] In one embodiment of the present application, the adjustment module 230 is specifically configured to adjust the images to be semantically segmented based on the image processing target parameter values ​​corresponding to each image output by the parameter preprocessing network, thereby obtaining multiple target images. The adjustment methods include, but are not limited to, adjusting image processing parameter values ​​such as brightness, contrast, and saturation.

[0074] It can be understood that the multiple target images here are images that need to be semantically segmented, that is, images that need to be predicted.

[0075] In one embodiment of the present application, the semantic segmentation module 240 is specifically configured to input the plurality of target images into the pre-trained semantic segmentation model, thereby performing image semantic segmentation and obtaining corresponding semantic segmentation results. Adjusting the image processing target parameter values ​​output by the preset parameter preprocessing network before inputting them into the semantic segmentation model can improve the semantic segmentation prediction effect.

[0076] It can be understood that the above-mentioned device for automatically learning network preprocessing parameters can implement the various steps of the method for automatically learning network preprocessing parameters provided in the aforementioned embodiment. The relevant explanations about the method for automatically learning network preprocessing parameters are applicable to the device for automatically learning network preprocessing parameters and will not be repeated here.

[0077] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 4 At the hardware level, the electronic device includes a processor and, optionally, an internal bus, a network interface, and memory. The memory may include internal memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for its services.

[0078] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0079] The memory is used to store programs. Specifically, the program may include program code, which includes computer operating instructions. The memory may include internal memory and non-volatile memory, and provides instructions and data to the processor.

[0080] The processor reads the corresponding computer program from the non-volatile memory into the internal memory and then runs it, forming a device that automatically learns the network preprocessing parameters at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0081] A semantic segmentation model is obtained by training randomly constructed image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each picture, and the image processing parameter items include image processing parameter values;

[0082] Input multiple images into the preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image;

[0083] Adjusting the pictures to be semantically segmented to obtain a plurality of target pictures according to the image processing target parameter value corresponding to each picture;

[0084] The multiple target images are input into the pre-trained semantic segmentation model to perform image semantic segmentation.

[0085] The above application Figure 1 The method performed by the apparatus for automatically learning network preprocessing parameters disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in the processor or by software instructions. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of this application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in conjunction with the embodiments of this application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0086] The electronic device may also perform Figure 1 The method for executing the device for automatically learning network preprocessing parameters in the embodiment of the present invention is realized. Figure 1 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0087] The embodiment of the present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by an electronic device including multiple application programs, can enable the electronic device to execute Figure 1 The method performed by the apparatus for automatically learning network preprocessing parameters in the embodiment shown is specifically used to perform:

[0088] A semantic segmentation model is obtained by training randomly constructed image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each picture, and the image processing parameter items include image processing parameter values;

[0089] Input multiple images into the preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image;

[0090] Adjusting the pictures to be semantically segmented to obtain a plurality of target pictures according to the image processing target parameter value corresponding to each picture;

[0091] The multiple target images are input into the pre-trained semantic segmentation model to perform image semantic segmentation.

[0092] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0094] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0096] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0097] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0098] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0099] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0100] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for automatically learning network preprocessing parameters, wherein: For a semantic segmentation model, the method includes: A semantic segmentation model is obtained by training randomly constructed image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each picture, and the image processing parameter items include image processing parameter values; Inputting multiple images into a preset parameter preprocessing network to obtain an image processing target parameter value corresponding to each image; the network structure of the preset parameter preprocessing network includes at least two fully connected layers; Input multiple images into the preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image, including: Multiple images are input into a preprocessing network with preset parameters, and processed sequentially through a two-dimensional normalization layer, a first fully connected layer, a one-dimensional normalization layer, an activation layer, and a second fully connected layer to obtain the image processing target parameter value corresponding to each image; Input multiple images into the preset parameter preprocessing network to obtain the image processing target brightness parameter value, target contrast parameter value and target saturation parameter value corresponding to each image; Adjusting the pictures to be semantically segmented to obtain multiple target pictures according to the image processing target brightness parameter value, target contrast parameter value, and target saturation parameter value corresponding to each picture; The multiple target images are input into the pre-trained semantic segmentation model to perform image semantic segmentation.

2. The method according to claim 1, wherein: The preset parameter preprocessing network is obtained through deep learning model training.

3. The method according to claim 2, wherein: After the semantic segmentation model is obtained by training the randomly constructed image processing parameter items, the following steps are further included: The semantic segmentation network part does not perform back propagation when the preset parameter preprocessing network trains the image processing target parameter value corresponding to each picture.

4. The method according to claim 1, wherein: The method further includes: performing data normalization processing on the image processing parameter value so that the image processing parameter value is within a preset range; The preset range interval is the same as the range interval of the random value corresponding to the randomly constructed image processing parameter item in the semantic segmentation model.

5. A device for automatically learning network preprocessing parameters, wherein: For a semantic segmentation model, the apparatus comprises: A semantic model training module, configured to train a semantic segmentation model by randomly constructing image processing parameter items, wherein the image processing parameter items are used to adjust the display effect of each image, and the image processing parameter items include image processing parameter values; The preprocessing network module is used to input multiple images into the preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image; The network structure of the preset parameter preprocessing network includes at least two fully connected layers; Input multiple images into the preset parameter preprocessing network to obtain the image processing target parameter value corresponding to each image, including: Multiple images are input into a preprocessing network with preset parameters, and processed sequentially through a two-dimensional normalization layer, a first fully connected layer, a one-dimensional normalization layer, an activation layer, and a second fully connected layer to obtain the image processing target parameter value corresponding to each image; Input multiple images into the preset parameter preprocessing network to obtain the image processing target brightness parameter value, target contrast parameter value and target saturation parameter value corresponding to each image; An adjustment module is used to adjust the image to be semantically segmented to obtain multiple target images according to the image processing target brightness parameter value, target contrast parameter value, and target saturation parameter value corresponding to each image; The semantic segmentation module is used to input the multiple target images into the pre-trained semantic segmentation model to perform image semantic segmentation.

6. An electronic device comprising: processor; as well as A memory arranged to store computer executable instructions, which when executed cause the processor to perform the method of any one of claims 1 to 4.

7. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of application programs, causes the electronic device to execute the method according to any one of claims 1 to 4.

Images