Object operation method and device, computer device, and computer storage medium

Abstract

The application discloses an object operation method and device, computer equipment and a computer storage medium, and belongs to the technical field of data processing. The method comprises the following steps: obtaining an object to be operated; inputting the object to be operated into a target model, the target model being a trained neural network model, and at least one parameter group in the target model being obtained through a preset mode; and obtaining an operation result output by the target model. The application inputs the object to be operated into the target model, and processes the object to be operated by the target model to output the operation result. Since the target model is a trained neural network model, the object library does not need to be relied on during processing, the problem that the processing success rate of the object operation method in the related art depends on the size of the object library and leads to low flexibility of the object operation method is solved, and the effect of improving the flexibility of the object operation method is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an object operation method and device, computer equipment and computer storage medium. BACKGROUND

[0002] An object operation method is a method for performing various operations on an object. The method can process and identify various objects such as images, sounds, signals, etc. to obtain an operation result.

[0003] In an object operation method, the similarity between a to-be-operated object and objects in an object library is compared. The object library includes a plurality of objects and an operation result corresponding to each object. If an object in the object library has a similarity greater than a specified value with the to-be-operated object, the operation result corresponding to the object in the object library is determined as the operation result of the to-be-operated object. For example, the to-be-operated object is an image, and the operation result corresponding to the image in the object library is a classification result corresponding to the content of the image.

[0004] However, the success rate of the above object operation method depends on the size of the object library, resulting in low flexibility of the object operation method. SUMMARY

[0005] Embodiments of the present application provide an object operation method, device, computer equipment and computer storage medium. The technical solution is as follows:

[0006] According to an aspect of an embodiment of the present application, an object operation method is provided, the method comprising:

[0007] obtaining a to-be-operated object;

[0008] inputting the to-be-operated object into a target model, the target model being a trained neural network model, and at least one parameter group in the target model being obtained by a preset manner;

[0009] obtaining an operation result output by the target model;

[0010] The preset manner includes: obtaining a sample parameter set corresponding to a first parameter group of the target model, the sample parameter set including a plurality of sample parameter groups, performing a plurality of iteration processes on the sample parameter set, obtaining a target parameter group based on the sample parameter set after the plurality of iteration processes, and determining the target parameter group as the first parameter group. Once the iteration process includes: obtaining four to-be-determined parameter groups of two sample parameter groups in the sample parameter set in a plurality of optimization directions, and replacing one sample parameter in the two sample parameters with a to-be-determined parameter having the smallest loss value in the four to-be-determined parameters.

[0011] Optionally, the sample parameter set is iteratively processed to obtain an iteration-processed sample parameter set.

[0012] The sample parameter set is iteratively processed to obtain an iteration-processed sample parameter set.

[0013] In response to the preset iteration termination condition not being reached, the iteration-processed sample parameter set is subjected to next iteration processing.

[0014] In response to the preset iteration termination condition being reached, the target parameter group is obtained based on the iteration-processed sample parameter set.

[0015] Optionally, the number of sample parameter groups in the sample parameter set is m+1, and the m+1 sample parameter groups are w n , w n+1 , w n+2 , …, and w n+m , where n is an integer greater than or equal to 0, and m is an integer greater than 2.

[0016] The four undetermined parameter groups of two sample parameter groups in the sample parameter set in multiple optimization directions are obtained, including:

[0017] The four undetermined parameter groups are obtained through a preset formula, and the preset formula includes:

[0018] w x = w n+1 + s*(w n+1 -w n ), where s is greater than 0.

[0019] w x+1 = w n+1 + 2s*(w n+2 -w n ).

[0020] w x+2 = w n+1 + u*(w n -w n+1 ), where u is greater than 0 and less than 1.

[0021] w x+3 = w n+1 + s*(w n+1 -w n ).

[0022] w n , w n+2 , w n+3 , and w n+4The four sets of parameters to be determined are: x is an integer greater than 0, and s and u are preset coefficients.

[0023] Optionally, replacing one of the two sample parameters with the undetermined parameter that has the smallest loss value among the four undetermined parameter groups includes:

[0024] In response to satisfying the first formula L n >L x L x ≥L x+1 Remove w from the sample parameter set n and the w x+1 w is determined to be in the set of sample parameters n+m+1 ;

[0025] In response to satisfying the second formula L n >L x L x <L x+1 Remove w from the sample parameter set n and the w x w is determined to be in the set of sample parameters n+m+1 ;

[0026] In response to satisfying the third formula L n ≤L x L x >L x+2 Remove w from the sample parameter set n and the w x+2 w is determined to be in the set of sample parameters n+m+1 ;

[0027] Since the first formula, the second formula, and the third formula are all not satisfied, w in the sample parameter set is removed. n and the w x+3 w is determined to be in the set of sample parameters n+m+1 ;

[0028] The L n For the w n The loss value, the L x For the w x The loss value, the L x+1 For the w x+1 The loss value, the L x+2 For the w x+2 The loss value.

[0029] Optionally, in response to reaching the preset iteration termination condition, obtaining the target parameter set based on the iteratively processed sample parameter set includes:

[0030] In response to the achievement of the preset iteration termination condition, the first sample parameter group with the smallest loss value in the sample parameter set after iteration is determined;

[0031] Obtain the mean sample parameter group from multiple sample parameter groups in the sample parameter set after iterative processing;

[0032] In response to the fact that the loss value of the first sample parameter group is less than the loss value of the mean sample parameter group, the first sample parameter group is determined to be the target sample parameter group;

[0033] In response to the loss value of the first sample parameter group being greater than the loss value of the mean sample parameter group, the mean sample parameter group is determined to be the target sample parameter group.

[0034] Optionally, after obtaining the sample parameter set after iterative processing, the method further includes:

[0035] In response to the number of iterations reaching a specified value, the preset iteration termination condition is determined;

[0036] If the number of iterations has not reached the specified value, it is determined that the preset iteration termination condition has not been met.

[0037] Optionally, after obtaining the sample parameter set after iterative processing, the method further includes:

[0038] Obtain the undetermined sample parameter group corresponding to the sample parameter set after the iterative processing. The undetermined sample parameter group is the mean sample parameter group of multiple sample parameter groups in the sample parameter set, or the undetermined sample parameter group is the sample parameter group with the smallest loss value in the sample parameter set.

[0039] In response to the loss value of the undetermined sample parameter group being less than or equal to a specified loss value, it is determined that the preset iteration termination condition has been met;

[0040] If the loss value of the undetermined sample parameter group is greater than the specified loss value, it is determined that the preset iteration termination condition has not been met.

[0041] Optionally, in response to reaching the preset iteration termination condition, obtaining the target parameter set based on the iteratively processed sample parameter set includes:

[0042] In response to the achievement of the preset iteration termination condition, the mean sample parameter group of multiple sample parameter groups in the sample parameter set after iteration processing is obtained.

[0043] The mean sample parameter set is determined as the target sample parameter set.

[0044] Optionally, in response to reaching the preset iteration termination condition, obtaining the target parameter set based on the iteratively processed sample parameter set includes:

[0045] In response to the achievement of the preset iteration termination condition, the first sample parameter group with the smallest loss value in the sample parameter set after iteration is obtained;

[0046] The first sample parameter group is determined as the target sample parameter group.

[0047] Optionally, before obtaining the four sets of parameters to be determined by a preset formula, the method further includes:

[0048] The w corresponding to the first parameter group is obtained sequentially. n The w n+1 The w n+2 and the w n+3 .

[0049] Optionally, the object to be operated on includes image data, sound data, and signal data.

[0050] According to another aspect of the embodiments of this application, an object manipulation device is provided, the object manipulation device comprising:

[0051] The object retrieval module is used to retrieve the object to be operated on.

[0052] An input module is used to input the object to be operated on into a target model, wherein the target model is a trained neural network model, and at least one set of parameters in the target model is obtained through a preset method;

[0053] The result acquisition module is used to obtain the operation results output by the target model.

[0054] The preset method includes: obtaining a sample parameter set corresponding to the first parameter set of the target model, the sample parameter set including multiple sample parameter sets; performing multiple iterations on the sample parameter set; obtaining a target parameter set based on the sample parameter set after multiple iterations; and determining the target parameter set as the first parameter set. Each iteration includes: obtaining four undetermined parameter sets of two sample parameter sets in multiple optimization directions; and replacing one of the two sample parameters with the undetermined parameter with the smallest loss value among the four undetermined parameter sets.

[0055] Optionally, the object manipulation device further includes:

[0056] The first iteration module is used to iteratively process the sample parameter set to obtain the iteratively processed sample parameter set;

[0057] The second iteration module is used to perform the next iteration on the sample parameter set after the iteration processing in response to the failure to meet the preset iteration termination condition;

[0058] The target acquisition module is used to acquire the target parameter set based on the sample parameter set after the iterative processing in response to the achievement of the preset iteration termination condition.

[0059] According to another aspect of the embodiments of this application, a computer device is provided, the computer device including a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, the at least one program, the code set or instruction set being loaded and executed by the processor to implement the object operation method as described above.

[0060] According to another aspect of the embodiments of this application, a non-transient computer storage medium is provided, wherein the computer storage medium stores at least one instruction, at least one program, code set or instruction set, wherein the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to implement the object operation method as described above.

[0061] A computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method described above.

[0062] The beneficial effects of the technical solutions provided in this application include at least the following:

[0063] By inputting the object to be operated on into the target model, and having the target model process the object to be operated on to output the operation result, since the target model is a trained neural network model, it can operate without relying on the object library. This solves the problem in related technologies where the success rate of object operation methods depends on the size of the object library, resulting in low flexibility of the object operation methods, and achieves the effect of improving the flexibility of object operation methods.

[0064] Furthermore, since at least one parameter set in the aforementioned target model is obtained through a preset method, and this preset method optimizes the parameter set using forward propagation, the computational load of parameter optimization is reduced, and the optimization speed is increased. This allows for faster acquisition of the target model for processing the object to be processed. In other words, the overall processing speed for the object to be operated on can be improved. Attached Figure Description

[0065] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0066] Figure 1 This is a schematic diagram of an object operating system provided in an embodiment of this application;

[0067] Figure 2 This is a flowchart illustrating an object manipulation method according to an embodiment of this application;

[0068] Figure 3 This is a flowchart of another object manipulation method provided in an embodiment of this application;

[0069] Figure 4 This is a flowchart of an iterative processing method according to an embodiment of this application;

[0070] Figure 5 This document illustrates a flowchart of an embodiment of the present application for obtaining a target parameter set based on a sample parameter set after iterative processing;

[0071] Figure 6 This is a two-dimensional contour map of an iterative process for parameter optimization in an embodiment of this application;

[0072] Figure 7 This is a structural block diagram of an object manipulation device provided in an embodiment of this application.

[0073] The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0075] The object manipulation methods provided in this application can be applied to object operating systems, such as... Figure 1 As shown, Figure 1 This is a schematic diagram of an object operating system provided in an embodiment of this application. The object operating system may include at least one of a server and a terminal. Figure 1Taking the object's operating system as an example, which includes a server and a terminal (but is not limited to this), the object's operating system can be used to process the object to be operated on. When the object's operating system includes a server 11 and a terminal 12, a wired connection and / or a wireless connection can be established between the server 11 and the terminal 12.

[0076] The server 11 may include a single server or a server cluster, and the terminal 12 may include a desktop computer, a laptop computer, a smartphone, or other smart wearable devices.

[0077] The object manipulation method provided in this application embodiment may include a model optimization process and an object manipulation process. Both processes may be implemented in server 11, or both processes may be implemented in terminal 12, or one process may be implemented in server 11 and the other process may be implemented in terminal 12. For example, the model optimization process may be implemented in server 11 and the object manipulation process may be implemented in terminal 12. This application embodiment does not impose any limitations on this.

[0078] The target model involved in the embodiments of this application can be a trained neural network model. A neural network (NN) model is a complex network model formed by a large number of interconnected processing units (called neurons). It reflects many fundamental characteristics of human brain function and is a highly complex nonlinear dynamic learning system. Neural network models possess large-scale parallelism, distributed storage and processing, self-organization, self-adaptation, and self-learning capabilities, making them suitable for handling information processing problems that require simultaneous consideration of many factors and conditions, and which are imprecise and fuzzy.

[0079] Neural network models are trained before application to improve their accuracy. During training, the parameters of the neural network model are optimized. A common optimization method is to use the backpropagation algorithm to calculate the gradient of the parameters. This method obtains the model's predicted values ​​through forward propagation, then obtains the gradient of the parameters through the backpropagation algorithm of the error. The parameters are then updated in the descent direction and proportion indicated by the gradient, iterating step by step to obtain the optimized parameters.

[0080] However, since the backpropagation algorithm mentioned above requires gradient calculation, which consumes a lot of computing resources, it will seriously affect the training speed of the model and place high demands on the computing power of the equipment used to train the model. All of these factors restrict the application of neural network models in object manipulation methods.

[0081] The object manipulation method provided in this application obtains four undetermined parameter groups from two sample parameter groups in multiple optimization directions within a sample parameter set. One of the two sample parameter groups is replaced by the undetermined parameter group with the smallest loss value among the four undetermined parameter groups. This achieves iteration of the parameter groups. This forward propagation method eliminates the need to calculate gradients, thereby reducing the computational load during parameter optimization. This improves the model's training speed and reduces the computational requirements of the device used to train the model, making it easier to apply neural network models to object manipulation methods.

[0082] Figure 2 This is a flowchart illustrating an object manipulation method according to an embodiment of this application. The object manipulation method may include the following steps:

[0083] Step 201: Obtain the object to be operated on.

[0084] Step 202: Input the object to be operated on into the target model. The target model is a trained neural network model, and at least one set of parameters in the target model is obtained through a preset method. The target model is used to perform recognition or processing operations on the object to be operated on.

[0085] Step 203: Obtain the operation results output by the target model.

[0086] The preset method includes: obtaining the sample parameter set corresponding to the first parameter set of the target model, the sample parameter set including multiple sample parameter sets, performing multiple iterations on the sample parameter set, obtaining the target parameter set based on the sample parameter set after multiple iterations, and determining the target parameter set as the first parameter set. One iteration process includes: obtaining four undetermined parameter sets of two sample parameter sets in multiple optimization directions, and replacing one of the two sample parameters with the undetermined parameter with the smallest loss value among the four undetermined parameter sets.

[0087] In summary, the object manipulation method provided in this application involves inputting the object to be manipulated into a target model, which then processes the object to be manipulated and outputs the operation result. Since the target model is a trained neural network model, it can process the object without relying on an object library, thus solving the problem in related technologies where the success rate of object manipulation methods depends on the size of the object library, resulting in low flexibility of the object manipulation method. This achieves the effect of improving the flexibility of the object manipulation method.

[0088] Furthermore, since at least one parameter set in the aforementioned target model is obtained through a preset method, and this preset method optimizes the parameter set using forward propagation, the computational load of parameter optimization is reduced, and the optimization speed is increased. This allows for faster acquisition of the target model for processing the object to be processed. In other words, the overall processing speed for the object to be operated on can be improved.

[0089] It should be noted that in the object operation method of this application embodiment, the target model is used to perform identification or processing operations on the object to be operated on. The identification operation can refer to the operation of identifying the object to be operated on to obtain an identification result, while the processing operation can refer to the operation of processing some or all of the data of the object to be operated on to obtain a processed object (the object to be operated on can be various types of data for the executing body of the object operation method, and the processing operation on the object to be operated on can include data processing operations). Specifically, the object to be operated on can be various types of data such as images, sounds, and signals. The results of the recognition and processing operations performed by the target model will differ depending on the type of object to be operated on. For example, when the object to be operated on is image data, the processing operations performed by the target model on the image data may include image restoration, enhancement, and adjustment, while the recognition operations performed on the image data may include recognizing objects, people, and text in the image data. When the object to be operated on is sound data, the processing operations performed by the target model on the sound data may include adjustment and editing, while the recognition operations performed on the sound data may include recognizing voiceprint information and language information (such as converting sound into text). When the object to be operated on is signal data, the processing and recognition operations performed on the signal data may include both processing and recognition of the signal data.

[0090] Figure 3 This is a flowchart of another object manipulation method provided in an embodiment of this application. This embodiment uses the application of this method in a server as an example for illustration. The object manipulation method may include the following steps:

[0091] Step 301: Sequentially obtain multiple sample parameter sets from the sample parameter set corresponding to the first parameter set of the target model.

[0092] The object manipulation method provided in this application embodiment can include a process of optimizing parameter values ​​in a target model, and a process of performing object manipulations through the target model. The target model can include at least one parameter group; this application embodiment uses the optimization of the first parameter value as an example for illustration.

[0093] During the optimization of the first parameter value, the server can obtain multiple sample parameter sets from the sample parameter set corresponding to the first parameter set in sequence. Based on the order of acquisition, these multiple sample parameter sets will also have an order, which can play a corresponding role in subsequent iterative processing.

[0094] For example, the number of sample parameter groups in the sample parameter set is 4, and the 4 sample parameter groups are w n w n+1 w n+2 and w n+3 Where n is an integer greater than 0. In this embodiment of the application, the initial set of sample parameters can be obtained by random initialization, for example, by initializing the parameter set with Gaussian distribution data to obtain the initial set of sample parameters.

[0095] Step 302: Iterate through the sample parameter set to obtain the iteratively processed sample parameter set.

[0096] This iterative process is a method of optimizing a set of sample parameters. It can be used to reduce the overall loss value of multiple sets of sample parameters in the sample parameter set.

[0097] For example, such as Figure 4 As shown, Figure 4 This is a flowchart of an iterative processing method according to an embodiment of this application, wherein one iteration may include the following steps:

[0098] Sub-step 3021: Obtain four undetermined parameter groups from two sample parameter groups in multiple optimization directions in the sample parameter set.

[0099] During each iteration, the server can select two sample parameter groups from the sample parameter set and obtain four undetermined parameter groups for multiple optimization directions from these two sample parameter groups. This is a forward propagation optimization method. When selecting parameters, the server can choose the first two sample parameter groups according to their order in the sample parameter set, that is, the first and second sample parameter groups in sequence.

[0100] In one exemplary embodiment, the number of sample parameter groups in the sample parameter set is m+1, and the m+1 sample parameter groups are w. n w n+1 w n+2 ···w n+m n is an integer greater than or equal to 0, and m is an integer greater than 2.

[0101] The server can obtain four sets of undetermined parameters through a preset formula; these four sets of undetermined parameters are w. n and wn+1 These two parameters represent four sets of undetermined parameters across multiple optimization directions.

[0102] The preset formulas include:

[0103] w x =w n+1 +s*(w n+1 -w n ), s > 0;

[0104] w x+1 =w n+1 +2s*(w n+2 -w n );

[0105] w x+2 =w n+1 +u*(w n -w n+1 ), u is greater than 0 and less than 1;

[0106] w x+3 =w n+1 +s*(w n+1 -w n );

[0107] w n w n+2 w n+3 and w n+4 There are four sets of parameters to be determined, where x is an integer greater than 0, and s and u are preset coefficients.

[0108] Sub-step 3022: Replace one of the two sample parameters with the undetermined parameter with the smallest loss value from the four undetermined parameter groups.

[0109] When implementing sub-step 3022, one approach may include:

[0110] In response to satisfying the first formula L n >L x L x ≥L x+1 Remove w from the sample parameter set n and w x+1 w is determined to be in the sample parameter set n+m+1 ;

[0111] In response to satisfying the second formula L n >L x L x <L x+1 Remove w from the sample parameter set n and w x w is determined to be in the sample parameter set n+m+1 ;

[0112] In response to satisfying the third formula L n ≤L x L x >L x+2 Remove w from the sample parameter set n and w x+2 w is determined to be in the sample parameter set n+m+1 ;

[0113] Since the first, second, and third formulas are not satisfied, w in the sample parameter set is removed. n and w x+3 w is determined to be in the sample parameter set n+m+1 ;

[0114] Among them, L n For w n The loss value, L x For w x The loss value, L x+1 For w x+1 The loss value, L x+2 For w x+2 The loss value.

[0115] It should be noted that since the application of the above four sets of conditions is mutually exclusive, in most cases, it is not necessary to perform four judgments and corresponding calculations. In most cases, only the first two judgments and the corresponding techniques are required.

[0116] Among them, the loss value L x+i =loss(y truth ,f(s;w x+i i = 0, 1, 2, 3, s is the input of the target model, y truth Given the true value corresponding to input s, f(s; w) x+i ) is the function corresponding to the target model.

[0117] Step 303: Determine whether the preset iteration termination condition has been met. If the preset iteration termination condition has been met, proceed to step 304. If the preset iteration termination condition has not been met, proceed to step 302.

[0118] The server can determine whether the preset iteration termination condition has been met after each iteration is completed.

[0119] In this embodiment of the application, the iteration termination condition may include a variety of conditions, and the server may terminate the iteration process when one of the iteration termination conditions is met.

[0120] The first method for determining the iteration termination condition includes:

[0121] 1) In response to the number of iterations reaching a specified value, determine that the preset iteration termination condition has been met;

[0122] 2) In response to the fact that the number of iterations has not reached the specified value, it is determined that the preset iteration termination condition has not been met.

[0123] In this case, the iteration terminates when the number of iterations reaches a specified value, which can be set in advance.

[0124] The second method for determining the iteration termination condition includes:

[0125] 1) Obtain the set of undetermined sample parameters corresponding to the set of sample parameters after iterative processing.

[0126] The undetermined sample parameter set is either the mean sample parameter set of multiple sample parameter sets in the sample parameter set, or the undetermined sample parameter set is the sample parameter set with the smallest loss value in the sample parameter set.

[0127] The mean sample parameter set can be the mean of multiple sample parameter sets in the sample parameter set after the current iteration. The mean can be an arithmetic mean or other types of averages, and this application embodiment does not limit this.

[0128] Mean sample parameter group loss value

[0129] The set of sample parameters w with the smallest loss value in the sample parameter set. i Loss value:

[0130] l = mi n Loss[y truth ,f(s;w i )];

[0131] The server can determine either the mean sample parameter group or the sample parameter group with the minimum loss value as the undetermined sample parameter group, or it can determine the smaller of the mean sample parameter group and the sample parameter group with the minimum loss value as the undetermined sample parameter group. This application embodiment does not impose any restrictions on this.

[0132] 2) When the loss value of the undetermined sample parameter group is less than or equal to the specified loss value, the preset iteration termination condition is determined to be met;

[0133] When the loss value of the undetermined sample parameter group is less than or equal to the specified loss value, it indicates that the undetermined sample parameter group meets the condition, and the server can determine that the preset iteration termination condition has been reached.

[0134] 3) If the loss value of the undetermined sample parameter group is greater than the specified loss value, it is determined that the preset iteration termination condition has not been met.

[0135] When the loss value of the undetermined sample parameter group is greater than the specified loss value, it indicates that the undetermined sample parameter group does not meet the conditions, and the server can determine that the preset iteration termination condition has not been met.

[0136] If the preset iteration termination condition is not met, the server can re-execute step 302 to perform the next iteration.

[0137] Step 304: Obtain the target parameter set based on the sample parameter set after iterative processing.

[0138] When the preset iteration termination condition is met, the server can obtain the target parameter set based on the sample parameter set after iteration processing.

[0139] In this embodiment of the application, the server can obtain the target parameter set based on the iteratively processed sample parameter set in various ways, for example, such as Figure 5 As shown, Figure 5 The flowchart illustrating an embodiment of this application for obtaining a target parameter set based on an iteratively processed sample parameter set includes the following steps:

[0140] Sub-step 3041: Determine the first sample parameter group with the smallest loss value in the sample parameter set after iterative processing.

[0141] The method for obtaining the first sample parameter group with the minimum loss value can refer to the above sub-step 303, and will not be repeated here in the embodiments of this application.

[0142] Sub-step 3042: Obtain the mean sample parameter group of multiple sample parameter groups in the sample parameter set after iterative processing.

[0143] The method for obtaining the first sample parameter group with the minimum loss value can refer to the above sub-step 303, and will not be repeated here in the embodiments of this application.

[0144] Sub-step 3043: In response to the fact that the loss value of the first sample parameter group is less than the loss value of the mean sample parameter group, the first sample parameter group is determined as the target sample parameter group.

[0145] Sub-step 3044: In response to the loss value of the first sample parameter group being greater than the loss value of the mean sample parameter group, determine the mean sample parameter group as the target sample parameter group.

[0146] In other words, the server can determine the target sample parameter group as the parameter group with the smaller loss value among the first sample parameter group and the mean sample parameter group.

[0147] Another process for obtaining the target parameter set based on the sample parameter set after iterative processing may include:

[0148] 1) Obtain the first set of sample parameters with the smallest loss value from the set of sample parameters after iterative processing;

[0149] The method for obtaining the first sample parameter group with the minimum loss value can refer to the above sub-step 303, and will not be repeated here in the embodiments of this application.

[0150] 2) Determine the first sample parameter set as the target sample parameter set;

[0151] In this approach, the server can determine the first sample parameter set as the target sample parameter set.

[0152] Step 305: Determine the target parameter set as the first parameter set of the target model.

[0153] The target sample parameter set is the optimized sample parameter set. The server can determine the target parameter set as the first parameter set of the target model in order to optimize the parameters in the target model.

[0154] Once step 305 is completed, the optimization process for the target model is finished. The server can optimize the parameter set in the target model using the methods shown in steps 301 to 305.

[0155] Step 306: Obtain the object to be operated on.

[0156] The object to be operated on can be various types of data, such as image data, sound data, and signal data.

[0157] It should be noted that the type of the object to be operated on can be the same as the type of the target model. If the objects that the target model can handle have been determined, the server can also obtain the corresponding type of object to be operated on in this step.

[0158] For example, if the target model is a model for recognizing images, the object to be operated on in step 306 can be image data; if the target model is a model for processing sound, the object to be operated on in step 306 can be sound data.

[0159] Step 307: Input the object to be operated on into the target model.

[0160] Once the server obtains the object to be operated on, it can input the object into the target model.

[0161] Step 308: Obtain the operation results output by the target model.

[0162] The server can obtain the operation results output by the target model.

[0163] The object manipulation methods provided in this application can be applied to various models, such as the LeNet network model and the AlexNet network model.

[0164] The LeNet network model was initially proposed by Turing Award winner LeCun in the late 20th century. The input to the LeNet network is a binary image of a handwritten digit, with a size of 32 pixels by 32 pixels. The LeNet network model consists of two convolutional layers, two pooling layers, and three fully connected layers. After the last fully connected layer, a sigmoid function is added to give the network non-linear fitting capabilities. In one specific embodiment, the output of the LeNet network model is a 10-dimensional vector. This LeNet network model performs an image classification task. Each dimension of the 10-dimensional vector corresponds to one of the digits 0 to 9. When the value at the corresponding position in the vector is 1, it means that the image is classified as corresponding to the handwritten digit.

[0165] The convolutional and fully connected layers in the LeNet network model have parameter sets that can be optimized. In the model training process of related technologies, the backpropagation algorithm is generally used for parameter optimization. The backpropagation algorithm requires the use of the chain rule (the chain rule is a derivative rule in calculus used to find the derivative of a composite function, and it is a commonly used method in calculus differentiation operations) to solve for the gradient, which is time-consuming and computationally intensive.

[0166] The object operation method provided in this application optimizes parameters through forward propagation. It can be applied to the LeNet network model to optimize the parameter set in the LeNet network model. Since the method provided in this application optimizes the parameter set with low computational load and short time consumption, it can improve the optimization speed of the LeNet network model and facilitate the rapid optimization of the LeNet network model for image recognition.

[0167] The AlexNet network model can perform tasks such as image classification. It takes a three-channel color RGB image as input and outputs a multi-dimensional vector, where each dimension represents a specific category of the image. Therefore, the vector's dimension is related to the number of image categories.

[0168] The AlexNet network model has 5 convolutional layers, 3 pooling layers, and 3 fully connected layers. These convolutional and fully connected layers also have parameter sets that can be optimized. Therefore, the AlexNet network model can also have its parameter sets optimized using the methods provided in the embodiments of this application.

[0169] In summary, the object manipulation method provided in this application involves inputting the object to be manipulated into a target model, which then processes the object to be manipulated and outputs the operation result. Since the target model is a trained neural network model, it can process the object without relying on an object library, thus solving the problem in related technologies where the success rate of object manipulation methods depends on the size of the object library, resulting in low flexibility of the object manipulation method. This achieves the effect of improving the flexibility of the object manipulation method.

[0170] Furthermore, since at least one parameter set in the aforementioned target model is obtained through a preset method, and this preset method optimizes the parameter set using forward propagation, the computational load of parameter optimization is reduced, and the optimization speed is increased. This allows for faster acquisition of the target model for processing the object to be processed. In other words, the overall processing speed for the object to be operated on can be improved.

[0171] The optimization method for the parameter group provided in the embodiments of this application will be further described below.

[0172] In an exemplary embodiment, taking the parameter set to be optimized in the target model as a two-dimensional parameter set as an example, the parameter set is represented as [a,b]^T, and the number of sample parameter sets in the preset sample parameter set is 4, λ=1, ρ=0.5.

[0173] Please refer to Figure 6 , Figure 6 This is a two-dimensional contour map of an iterative process for parameter optimization in an embodiment of this application. Figure 6 The two concentric curves in the figure are contour lines representing the loss function values, describing the loss values ​​at different parameter mapping locations. Points A, B, C, and D in the figure represent the four initially acquired sample parameter sets, which together constitute the initial sample parameter set.

[0174] The process of the first iteration may include:

[0175] Take the parameters represented by points A and B, and apply the parameter w. A w B The calculation yields w A w B The corresponding four sets of undetermined sample parameters: w 01 =w E , Based on the magnitude of the loss value corresponding to each point in the graph (the closer to the center, the smaller the loss value), it can be concluded that... Established (l) v Let v represent the loss value (E, A, E1), and let the parameter w A Remove parameter w from the alternative optimization parameter group. EAdd to the sample parameter set.

[0176] At the end of the first iteration, the sample parameter set contains parameter sets corresponding to points B, C, D, and E.

[0177] In the second iteration, points B and C are selected, and after calculation, the parameter set w can be obtained. F (The calculation process is omitted here, and w is assumed to be...) F To determine the parameters that satisfy the conditions involved in step 302, add them to the sample parameter set.

[0178] After multiple iterations, from Figure 6 As can be seen, the loss function value corresponding to the parameter set gradually approaches the minimum point.

[0179] When the parameter update reaches the iteration termination condition, assume that there are points H, I, J, and K in the sample parameter set.

[0180] We can assume w k (That is, the parameter set corresponding to point K) has the minimum loss value, and w k The loss value is l.

[0181] Its average parameter is set as w z :

[0182]

[0183]

[0184] according to Figure 6 It can be seen that point K is located at a local minimum point in the parameter space. Therefore, we take w. k To obtain the optimal parameter set, w can be... k Deployed in the target model.

[0185] In the object operation method provided in this application embodiment, the method for optimizing the parameter set is a local minimum point solution optimization method (also known as a weighted walk algorithm). This method can satisfy the same precondition as the gradient descent method, that is, a differentiable convex function within the range of values ​​of the function being optimized.

[0186] Assume the optimal parameter is w * Then we have f′(w * )=0,f(w * )≤f(w), where f(w) is the loss function. Gradient descent requires calculating the first derivative f′(w) of the loss function f(w). The value of the function f′(w0) is the gradient of the original function. The negative direction of the gradient is the steepest direction of the decrease in function value. With the help of the first derivative, gradient descent makes the function value continuously decrease. When f′(w)→0, it is determined that the function is close to the minimum point.

[0187] According to the definition of gradient:

[0188]

[0189] Gradient descent controls the magnitude of parameter adjustment through the gradient value, and controls the direction of parameter adjustment through the sign of the gradient value. According to the definition of gradient, the sign of f′(w) depends on the sign of f(w+Δw)-f(w). The direction of gradient descent is the direction in which f(w+Δw)-f(w)<0.

[0190] The method proposed in this application calculates the value of the function f(x). As mentioned above, the function to be optimized is a convex function, therefore it has one and only one set of parameters w. * such that min f(w)=f(w) * The condition is met, and distance(w,w) holds true. * )∝f(w)-f(w * The weighted walk optimization algorithm initializes multiple sets of parameters and continuously updates the function values ​​of the parameters, causing the function value f(w) to continuously decrease, i.e., f(w) - f(w). * The distance(w,w) decreases continuously, thus making the distance(w,w) decrease. * As the parameter value decreases continuously and approaches a local minimum, the optimization of the parameter set in the objective function is achieved.

[0191] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein. For details not disclosed in the apparatus embodiments of this disclosure, please refer to the embodiments of the method disclosed herein.

[0192] Figure 7 This is a structural block diagram of an object manipulation device provided in an embodiment of this application. The object manipulation device 700 includes:

[0193] The object acquisition module 710 is used to acquire the object to be operated on.

[0194] The input module 720 is used to input the object to be operated into the target model, which is a trained neural network model, and at least one set of parameters in the target model is obtained through a preset method.

[0195] Result acquisition module 730 is used to obtain the operation results output by the target model;

[0196] The preset method includes: obtaining the sample parameter set corresponding to the first parameter set of the target model, the sample parameter set including multiple sample parameter sets, performing multiple iterations on the sample parameter set, obtaining the target parameter set based on the sample parameter set after multiple iterations, and determining the target parameter set as the first parameter set. One iteration process includes: obtaining four undetermined parameter sets of two sample parameter sets in multiple optimization directions, and replacing one of the two sample parameters with the undetermined parameter with the smallest loss value among the four undetermined parameter sets.

[0197] In summary, the object manipulation device provided in this application allows the object to be manipulated to be input into a target model, which then processes the object to be manipulated and outputs the operation result. Since the target model is a trained neural network model, it can process the object without relying on an object library, thus solving the problem that the success rate of object manipulation methods in related technologies depends on the size of the object library, resulting in low flexibility of the object manipulation method. This achieves the effect of improving the flexibility of the object manipulation method.

[0198] Furthermore, since at least one parameter set in the aforementioned target model is obtained through a preset method, and this preset method optimizes the parameter set using forward propagation, the computational load of parameter optimization is reduced, and the optimization speed is increased. This allows for faster acquisition of the target model for processing the object to be processed. In other words, the overall processing speed for the object to be operated on can be improved.

[0199] Optionally, the object manipulation device further includes:

[0200] The first iteration module is used to iteratively process the sample parameter set to obtain the iteratively processed sample parameter set.

[0201] The second iteration module is used to perform the next iteration on the sample parameter set after the iteration process in response to the failure to meet the preset iteration termination condition;

[0202] The target acquisition module is used to obtain the target parameter set based on the sample parameter set after iterative processing in response to the achievement of a preset iteration termination condition.

[0203] Optionally, the number of sample parameter groups in the sample parameter set is m+1, and the m+1 sample parameter groups are w n w n+1 w n+2 ···w n+m n is an integer greater than or equal to 0, and m is an integer greater than 2;

[0204] The object manipulation device also includes: a parameter acquisition module, used for:

[0205] Four sets of parameters to be determined are obtained through preset formulas, including:

[0206] w x =w n+1 +s*(w n+1 -w n ), s > 0;

[0207] w x+1 =w n+1 +2s*(w n+2 -w n );

[0208] w x+2 =w n+1 +u*(w n -w n+1 ), u is greater than 0 and less than 1;

[0209] w x+3 =w n+1 +s*(w n+1 -w n );

[0210] w n w n+2 w n+3 and w n+4 There are four sets of parameters to be determined, where x is an integer greater than 0, and s and u are preset coefficients.

[0211] Optionally, the object manipulation device further includes: a parameter replacement module, used for:

[0212] In response to satisfying the first formula L n >L x L x ≥L x+1 Remove w from the sample parameter set n and w x+1 w is determined to be in the sample parameter set n+m+1 ;

[0213] In response to satisfying the second formula L n >L x L x <L x+1 Remove w from the sample parameter set n and w x w is determined to be in the sample parameter set n+m+1 ;

[0214] In response to satisfying the third formula L n ≤L x L x >L x+2Remove w from the sample parameter set n and w x+2 w is determined to be in the sample parameter set n+m+1 ;

[0215] Since the first, second, and third formulas are not satisfied, w in the sample parameter set is removed. n and w x+3 w is determined to be in the sample parameter set n+m+1 ;

[0216] L n For w n The loss value, L x For w x The loss value, L x+1 For w x+1 The loss value, L x+2 For w x+2 The loss value.

[0217] Optionally, the object manipulation device further includes: a first target parameter group acquisition module, used for:

[0218] In response to the achievement of the preset iteration termination condition, the first set of sample parameters with the smallest loss value is determined in the set of sample parameters after iteration processing;

[0219] Obtain the mean value of multiple sample parameter groups in the sample parameter set after iterative processing;

[0220] Since the loss value of the first sample parameter group is less than the loss value of the mean sample parameter group, the first sample parameter group is determined as the target sample parameter group.

[0221] In response to the fact that the loss value of the first sample parameter group is greater than the loss value of the mean sample parameter group, the mean sample parameter group is determined as the target sample parameter group.

[0222] Optionally, the object manipulation device further includes: a first iteration termination determination module, used for:

[0223] In response to the number of iterations reaching a specified value, the preset iteration termination condition is determined;

[0224] If the number of iterations has not reached the specified value, it is determined that the preset iteration termination condition has not been met.

[0225] Optionally, the object manipulation device further includes: a second iteration termination determination module, used for:

[0226] Obtain the undetermined sample parameter group corresponding to the sample parameter set after iterative processing. The undetermined sample parameter group is the mean sample parameter group of multiple sample parameter groups in the sample parameter set, or the undetermined sample parameter group is the sample parameter group with the smallest loss value in the sample parameter set.

[0227] When the loss value of the parameter group of the undetermined sample is less than or equal to the specified loss value, the preset iteration termination condition is determined to be met.

[0228] If the loss value of the undetermined sample parameter group is greater than the specified loss value, it is determined that the preset iteration termination condition has not been met.

[0229] Optionally, the object manipulation device further includes: a second target parameter set acquisition module, used for:

[0230] In response to reaching the preset iteration termination condition, the mean sample parameter set of multiple sample parameter sets in the sample parameter set after iteration processing is obtained.

[0231] The mean sample parameter set is determined as the target sample parameter set.

[0232] Optionally, the object manipulation device further includes: a third target parameter group acquisition module, used for:

[0233] Obtain the first set of sample parameters with the smallest loss value from the set of sample parameters after iterative processing;

[0234] The first sample parameter group is determined as the target sample parameter group.

[0235] Optionally, the object manipulation device further includes: a sequential acquisition module, used for:

[0236] Sequentially obtain the w corresponding to the first parameter group n w n+1 w n+2 and w n+3 .

[0237] Optionally, the objects to be operated on include image data, sound data, and signal data.

[0238] According to another aspect of the embodiments of this application, a computer device is provided, the computer device including a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, at least one program, code set or instruction set being loaded and executed by the processor to implement the object operation method as described above.

[0239] According to another aspect of the embodiments of this application, a non-transient computer storage medium is provided, wherein at least one instruction, at least one program, code set or instruction set is stored in the computer storage medium, and at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the object operation method as described above.

[0240] A computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method described above.

[0241] In this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0242] In this application, the term "at least one of A and B" merely describes the relationship between related objects, indicating that three relationships can exist. For example, "at least one of A and B" can represent: A existing alone, A and B existing simultaneously, and B existing alone. Similarly, "at least one of A, B, and C" indicates that seven relationships can exist, representing: A existing alone, B existing alone, C existing alone, A and B existing simultaneously, A and C existing simultaneously, C and B existing simultaneously, and A, B, and C existing simultaneously. Likewise, "at least one of A, B, C, and D" indicates that fifteen relationships can exist, representing: A existing alone, B existing alone, C existing alone, D existing alone, A and B existing simultaneously, A and C existing simultaneously, A and D existing simultaneously, C and B existing simultaneously, D and B existing simultaneously, C and D existing simultaneously, A, B, and C existing simultaneously, A, B, and D existing simultaneously, A, C, and D existing simultaneously, and A, B, C, and D existing simultaneously.

[0243] In this application, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. The term "multiple" refers to two or more unless otherwise expressly defined.

[0244] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0245] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0246] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0247] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An object operation method characterized by, The method comprises: acquiring a to-be-operated object, the to-be-operated object comprising image data, sound data and signal data; inputting the to-be-operated object into a target model, the target model being a trained neural network model, and at least one parameter group in the target model being obtained by a preset manner, the target model being used for performing an identification operation or a processing operation on the to-be-operated object; acquiring an operation result output by the target model; wherein the preset manner comprises: acquiring a sample parameter set corresponding to a first parameter group of the target model, the sample parameter set comprising a plurality of sample parameter groups, performing a plurality of iteration processes on the sample parameter set, acquiring a target parameter group based on the sample parameter set after the plurality of iteration processes, and determining the target parameter group as the first parameter group, one iteration process comprising: acquiring four to-be-determined parameter groups of two sample parameter groups in the sample parameter set in a plurality of optimization directions, and replacing one sample parameter in the two sample parameters with a to-be-determined parameter with a minimum loss value in the four to-be-determined parameter groups; The number of sample parameter groups in the sample parameter set is m+1, and the m+1 sample parameter groups are w n , w n+1 , w n+2 ,..., w n+m , n is an integer greater than or equal to 0, and m is an integer greater than 2. the acquiring of the four to-be-determined parameter groups of the two sample parameter groups in the sample parameter set in the plurality of optimization directions comprises: acquiring the four to-be-determined parameter groups by a preset formula, the preset formula comprising: w x = w n+1 + s*(w n+1 - w n ), s is greater than 0; w x+1 = w n+1 + 2s*(w n+2 - w n ); w x+2 = w n+1 + u * (w n - w n+1 ), u is greater than 0 and less than 1; w x+3 = w n+1 + s*(w n+1 - w n ); The w n , w n+2 , w n+3 , and w n+4 are the four groups of undetermined parameters, x is an integer greater than 0, and s and u are preset coefficients.

2. The method of claim 1, wherein, before the acquiring of the to-be-operated object, the performing of the plurality of iteration processes on the sample parameter set, the acquiring of the target parameter group based on the sample parameter set after the plurality of iteration processes comprises: performing an iteration process on the sample parameter set to obtain an iteration-processed sample parameter set; in response to a preset iteration termination condition not being reached, performing a next iteration process on the iteration-processed sample parameter set; in response to the preset iteration termination condition being reached, acquiring the target parameter group based on the iteration-processed sample parameter set.

3. The method of claim 1, wherein, the replacing of one sample parameter in the two sample parameters with a to-be-determined parameter with a minimum loss value in the four to-be-determined parameter groups comprises: in response to satisfying a first formula L n L x , L x ≥ L x+1 , removing w n from the set of sample parameters and determining w x+1 as w n+m+1 from the set of sample parameters; in response to satisfying a second formula L n L x L x L x+1 , removing w n from the set of sample parameters and determining w x as w n+m+1 from the set of sample parameters; In response to satisfying a third formula L n ≤ L x , L x > L x+2 , remove w n from the set of sample parameters and determine w x+2 as w n+m+1 from the set of sample parameters; remove w from the sample parameter set in response to none of the first formula, the second formula, and the third formula being satisfied n , and determine the w x+3 as the w n+m+1 in the sample parameter set The L n loss value of the w n The L x loss value of the w x The L x+1 loss value of the w x+1 The L x+2 loss value of the w x+2 loss value of the w 4. The method of claim 2, wherein, the acquiring of the target parameter group based on the iteration-processed sample parameter set in response to the preset iteration termination condition being reached comprises: in response to the preset iteration termination condition being reached, determining a first sample parameter group with a minimum loss value in the iteration-processed sample parameter set; acquiring a mean sample parameter group of a plurality of sample parameter groups in the iteration-processed sample parameter set; in response to the loss value of the first sample parameter group being smaller than the loss value of the mean sample parameter group, determining the first sample parameter group as the target sample parameter group; in response to the loss value of the first sample parameter group being greater than the loss value of the mean sample parameter group, determining the mean sample parameter group as the target sample parameter group.

5. The method of claim 2, wherein, after the obtaining of the iteration-processed sample parameter set, the method further comprises: in response to the number of iteration processes reaching a specified value, determining that the preset iteration termination condition is reached; in response to the number of iteration processes not reaching the specified value, determining that the preset iteration termination condition is not reached.

6. The method of claim 2, wherein, after the obtaining of the iteration-processed sample parameter set, the method further comprises: obtaining a to-be-determined sample parameter group corresponding to the sample parameter set after the iteration processing, the to-be-determined sample parameter group being a mean sample parameter group of multiple sample parameter groups in the sample parameter set, or the to-be-determined sample parameter group being a sample parameter group with a minimum loss value in the sample parameter set; in response to the loss value of the to-be-determined sample parameter group being less than or equal to a specified loss value, determining that the preset iteration termination condition is reached; in response to the loss value of the to-be-determined sample parameter group being greater than the specified loss value, determining that the preset iteration termination condition is not reached.

7. The method of claim 2, wherein, The method further comprises the following steps before obtaining the four to-be-determined parameter groups through the preset formula: The object operation apparatus comprises: an object obtaining module, configured to obtain a to-be-operated object, the to-be-operated object comprising image data, sound data, and signal data; 8. The method of claim 2, wherein, an input module, configured to input the to-be-operated object into a target model, the target model being a trained neural network model, and at least one parameter group in the target model being obtained through a preset manner, the target model being configured to perform an identification operation or a processing operation on the to-be-operated object; a result obtaining module, configured to obtain an operation result output by the target model; wherein the preset manner comprises: obtaining a sample parameter set corresponding to a first parameter group of the target model, the sample parameter set comprising multiple sample parameter groups, performing multiple iteration processes on the sample parameter set, obtaining a target parameter group based on the sample parameter set after the multiple iteration processes, and determining the target parameter group as the first parameter group, one iteration process comprising: obtaining four to-be-determined parameter groups of two sample parameter groups in the sample parameter set in multiple optimization directions, and replacing one sample parameter in the two sample parameters with a to-be-determined parameter with a minimum loss value in the four to-be-determined parameter groups; 9. The method of claim 1, wherein, The method further comprises the following steps before obtaining the four to-be-determined parameter groups through the preset formula: obtaining the w n , the w n+1 , the w n+2 , and the w n+3 .

10. An object operation device characterized by comprising: The preset formula comprises: The object operation apparatus further comprises: a first iteration module, configured to perform iteration processing on the sample parameter set to obtain a sample parameter set after iteration processing; ​ ​ The number of sample parameter groups in the sample parameter set is m+1, and the m+1 sample parameter groups are w n , w n+1 , w n+2 , w n+m , and n is an integer greater than or equal to 0, and m is an integer greater than 2. ​ ​ w x = w n+1 + s*(w n+1 - w n ), s is greater than 0; w x+1 = w n+1 + 2s*(w n+2 - w n ); w x+2 = w n+1 + u * (w n - w n+1 ), u is greater than 0 and less than 1; w x+3 = w n+1 + s*(w n+1 - w n ); The w n , w n+2 , w n+3 , and w n+4 are the four groups of undetermined parameters, x is an integer greater than 0, and s and u are preset coefficients.

11. The object manipulation device of claim 10, wherein, ​ ​ a second iteration module, configured to perform next iteration processing on the sample parameter set after the iteration processing in response to the preset iteration termination condition not being reached; an object acquisition module, configured to acquire the target parameter group based on the sample parameter set after the iteration processing in response to the preset iteration termination condition being reached.

12. A computer device, comprising: The computer device comprises a processor and a memory, and the memory stores at least one instruction, at least one program, a code set or an instruction set, which are loaded and executed by the processor to implement the object operation method according to any one of claims 1 to 9.

13. A non-transitory computer storage medium, comprising: The computer storage medium stores at least one instruction, at least one program, a code set or an instruction set, which are loaded and executed by the processor to implement the object operation method according to any one of claims 1 to 9.

Images