Humanoid entity terrorist valley effect evaluation method, system, device and medium

Abstract

According to the human-like entity terrorist valley effect evaluation method, system, device and medium provided by the invention, the face data set and the neural network based on the intermediary effect model are constructed, so that the evaluation accuracy of the affinity and anthropomorphic degree of the human-like entity is effectively improved, and the evaluation accuracy of the affinity and anthropomorphic degree of the human-like entity is improved. Therefore, the design of the human-like entity is optimized. The technical effect can reduce the negative influence of the terrorist valley effect on the user emotion experience, and enhance the acceptability and popularity rate of the human-like entity.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence evaluation technology, and in particular to a method, system, device and medium for evaluating the uncanny valley effect of a humanoid entity. Background Art

[0002] In the development of modern science and technology, humanoid entities (such as social robots, digital humans, virtual assistants, and virtual characters in video games) have gradually been integrated into people's daily lives. These humanoid entities imitate human morphology, voice, and behavior to provide a more natural and intuitive interactive experience. However, when the appearance and behavior of humanoid entities are close to humans but have not yet reached a completely realistic level, they may trigger the "uncanny valley effect". The uncanny valley effect describes the changing trend of people's emotional response to humanoid entities as their realism gradually increases. Specifically, when the realism of humanoid entities is low, people's positive emotions gradually increase as their realism increases. However, when its realism reaches a certain critical point, people's positive emotions reach a peak and then drop sharply, generating negative emotions such as fear and disgust. Only when the realism of humanoid entities is closer to the real level of humans will negative emotions weaken and people's positive emotions will gradually recover. This uncanny valley effect causes people to feel uncomfortable or even disgusted with humanoid entities, thus limiting their acceptance and popularity in practical applications.

[0003] Existing research on the uncanny valley effect has multiple deficiencies in measurement and evaluation techniques, including reliability issues of subjective evaluation methods, limitations of physiological and neural measurement methods, lack of in-depth understanding of artificial intelligence models, insufficient diversity of data sets, and lack of unified quantitative evaluation standards. These problems limit the comprehensive understanding and precise quantification of the uncanny valley effect, and new technologies and methods are urgently needed to overcome these challenges. Summary of the invention

[0004] The present disclosure provides a method, system, device and storage medium for evaluating the uncanny valley effect of a humanoid entity.

[0005] According to one aspect of the present disclosure, a method for evaluating the uncanny valley effect of a humanoid entity is provided, comprising:

[0006] Collecting facial image data of humanoid entities and constructing a facial dataset of humanoid entities; the facial dataset includes images of different degrees of anthropomorphism, and the facial features of the facial image data include at least geometric features, texture features, and hair features;

[0007] Soliciting perceptual evaluations of the facial data set, and collecting perceptual evaluation data of the humanoid entity; the perceptual evaluation data includes anthropomorphism perceptual evaluation data and affinity perceptual evaluation data, and the perceptual evaluation includes geometry, texture, hair anthropomorphism perceptual evaluation, overall anthropomorphism perception, and affinity perceptual evaluation;

[0008] Performing feature quantification analysis on the perception evaluation data and the facial features through a mediation effect model to obtain a quantified result;

[0009] Constructing an uncanny valley effect evaluation model for humanoid entities based on a pre-trained facial feature encoder and the quantization results;

[0010] The uncanny valley effect evaluation model of the humanoid entity is used to evaluate the facial image of the humanoid entity.

[0011] In the method according to some embodiments of the present disclosure, the collecting facial image data of the human-like entity includes:

[0012] By collecting and / or synthesizing facial image data of the humanoid entity; the types of the humanoid entity at least include robots, humanoid models, digital humans, computer-generated characters, real dolls, wax figures and real humans.

[0013] In the method according to some embodiments of the present disclosure, the step of constructing a facial dataset includes:

[0014] The facial image data of the humanoid entity is preprocessed by using a white background and combining high-resolution frontal portraits and neutral expression images to construct the facial dataset.

[0015] In the method according to some embodiments of the present disclosure, the step of performing feature quantification analysis on the perception evaluation data of the humanoid entity and the facial features through a mediation effect model to obtain a quantification result includes:

[0016] The anthropomorphism perception of the geometric features, the texture features and the hair features in the facial features is used as an independent variable, the overall anthropomorphism in the anthropomorphism perception evaluation data is used as a mediating variable, and affinity is used as a dependent variable;

[0017] Using multiple regression analysis to evaluate the first effect of the geometric features, the texture features and the hair features on the affinity, using multiple regression analysis to evaluate the second effect of the geometric features, the texture features and the hair features on the overall anthropomorphism, using univariate regression analysis to evaluate the third effect of the overall anthropomorphism on the affinity, and using the multiplication rule to calculate the indirect effect;

[0018] The Sobels test method was used to evaluate the significance of the indirect effect and determine the mediating effect;

[0019] A comprehensive regression model is constructed by combining the first influence, the second influence and the third influence to evaluate the total influence of the independent variable and the mediating variable on the dependent variable to obtain the quantitative result; wherein the quantitative result includes direct effect and indirect effect.

[0020] In the method according to some embodiments of the present disclosure, the step of training the facial feature encoder includes:

[0021] Extracting the geometric features, the texture features, and the hair features from a facial image of a humanoid entity;

[0022] During the training process, the output of the encoder is regularized by KL divergence, so that the facial feature encoder generates latent variables with distribution characteristics, wherein the KL divergence is used to measure the difference between the distribution generated by the training encoder and the standard normal distribution.

[0023] In the method according to some embodiments of the present disclosure, the step of constructing an uncanny valley effect evaluation model of a humanoid entity based on a pre-trained facial feature encoder and the quantization result includes:

[0024] Based on the quantization result, weighting processing is performed on the extracted geometric features, the texture features and the hair features to generate direct effect weighted features and indirect effect weighted features;

[0025] Dynamically fusing the direct effect weighted features and the indirect effect weighted features through an attention mechanism to obtain a fused feature;

[0026] The fused features are input into two parallel branches for predicting the degree of anthropomorphism and predicting affinity, so as to complete the construction of the uncanny valley effect evaluation model of the humanoid entity.

[0027] In the method according to some embodiments of the present disclosure, it also includes:

[0028] Introducing a multi-task loss function to optimize the uncanny valley effect evaluation model of the humanoid entity;

[0029] The definition of the multi-task loss function includes:

[0030] Anthropomorphism loss:

[0031]

[0032] Where N is the number of samples, M is i is the true anthropomorphism value.

[0033] Affinity loss:

[0034]

[0035] Among them, Y i is the true anthropomorphism value.

[0036] The total loss function is:

[0037] L=λ1L M +λ2L Y

[0038] Among them, λ1 and λ2 are weight hyperparameters that control the two loss terms.

[0039] According to another aspect of the present disclosure, a system for evaluating the uncanny valley effect of a humanoid entity is provided, comprising:

[0040] A data set establishment module, used to collect facial image data of humanoid entities and construct a facial data set of humanoid entities; the facial data set of humanoid entities includes images of different degrees of anthropomorphism, and the facial features of the facial image data include at least geometric features, texture features and hair features;

[0041] An evaluation module, used to collect perceptual evaluations of the facial data set, and collect perceptual evaluation data of the humanoid entity; the perceptual evaluation data includes anthropomorphism perceptual evaluation data and affinity perceptual evaluation data, and the perceptual evaluation includes geometry, texture, hair anthropomorphism perceptual evaluation, overall anthropomorphism perception and affinity perceptual evaluation;

[0042] A mediating effect model module, used to perform feature quantification analysis on the perception evaluation data and the facial features through a mediating effect model to obtain a quantified result;

[0043] A model building module, used to build an uncanny valley effect evaluation model of a humanoid entity based on a pre-trained facial feature encoder and the quantization result;

[0044] An evaluation module evaluates the facial image of the humanoid entity using the uncanny valley effect evaluation model of the humanoid entity.

[0045] According to another aspect of the present disclosure, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the above method when executing the computer program.

[0046] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are implemented.

[0047] This paper proposes a method for evaluating the uncanny valley effect of humanoid entities. By constructing a facial dataset (HumaniFace-1K) and a neural network based on a mediating effect model (HumaniNet framework), it effectively improves the accuracy of evaluating the affinity and anthropomorphism of humanoid entities, thereby optimizing the design of humanoid entities. This technical effect can reduce the negative impact of the uncanny valley effect on the user's emotional experience, enhance the acceptance and popularity of humanoid entities, and open up new possibilities for future artificial intelligence applications.

[0048] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0050] Figure 1 is a flowchart according to an exemplary embodiment of the present disclosure;

[0051] Figure 2 is a schematic diagram of an overview of the HumaniNet model framework according to an exemplary embodiment of the present disclosure;

[0052] Figure 3 is a block diagram of a system for evaluating the uncanny valley effect of a humanoid entity according to an exemplary embodiment of the present disclosure;

[0053] Figure 4 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0054] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0055] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0056] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0057] The existing measurement and evaluation methods of the uncanny valley effect have the following problems:

[0058] The insufficiency of existing datasets. The construction of datasets is an important basis for the study of the uncanny valley effect. The researchers constructed a dataset of images of humanoid entities with different degrees of anthropomorphism, invited subjects to evaluate and conduct experimental analysis. Current datasets for the study of the uncanny valley effect often have problems such as limited sample size, single sample type, and insufficient diversity. These datasets are usually unable to fully cover all types of entities from low to high anthropomorphism, resulting in limited generalization of research results. In addition, some datasets use image deformation technology to generate samples, which may produce unrealistic transition images. The weirdness of these images themselves may interfere with the experimental results and make it difficult to accurately reflect the true situation of the uncanny valley effect.

[0059] Subjective evaluation method: Subjective evaluation method is the most commonly used technical means in the study of the uncanny valley effect. Researchers use tools such as Likert scale, visual analog scale, semantic differential scale, etc. to collect data on the subjects' emotional reactions to humanoid entities. These scales quantify the intensity of the uncanny valley effect by evaluating the subjects' emotional dimensions such as comfort, weirdness, and pleasure. However, the subjective evaluation method is highly dependent on the subjects' self-reports and is easily affected by factors such as individual differences, cultural background, and social expectations, resulting in poor reliability and consistency of the data. In addition, there are structural problems in the design of existing scales. For example, the definition of some emotional dimensions is too single or the dimensions are highly correlated, which cannot fully capture the complexity of the uncanny valley effect. The influence of psychometric noise and social idealized answers may also further weaken the validity of the evaluation results.

[0060] Physiological and behavioral signal measurement methods. In order to obtain more objective data, researchers have used eye tracking, functional magnetic resonance imaging (fMRI), electroencephalography (EEG) and other technologies to record the physiological and neural responses of humans when facing humanoid entities. These measurement methods attempt to reveal the neural mechanisms behind the uncanny valley effect through physiological indicators. However, it is difficult to obtain physiological and behavioral signals from subjects. Although fMRI provides high spatial resolution images of brain activity, its temporal resolution is low and it cannot capture rapidly changing neural responses in real time. In addition, the BOLD signal of fMRI is an indirect measurement of neural activity and may not accurately reflect immediate neural activity. Although EEG has a high temporal resolution, its spatial resolution is low, making it difficult to accurately locate specific brain activity areas. These technologies also have problems such as expensive equipment, complex operation, and difficulty in widespread application in large-scale studies.

[0061] Artificial intelligence and machine learning methods, with the development of deep learning and artificial intelligence technology, researchers have begun to use neural network models (such as CLIP, SVM, etc.) to simulate and predict the uncanny valley effect. By inputting a large amount of image data of humanoid entities, these models can learn and predict the emergence of the uncanny valley effect, thus providing a new technical means for the study of the uncanny valley effect. However, when simulating the uncanny valley effect, existing artificial intelligence models mainly rely on large-scale data training, but lack a deep understanding of the complex emotional responses of humans. This may cause the model to be biased when predicting human perceptual responses to humanoid entities. In addition, the performance of artificial intelligence models is highly dependent on the quality of the training data set. The insufficient number and diversity of samples in the existing humanoid entity data set may limit the generalization ability of the model, resulting in inaccurate prediction results.

[0062] It is for the above reasons that Figure 1 This embodiment provides a method for evaluating the uncanny valley effect of a humanoid entity, comprising:

[0063] S101, collecting facial image data of human-like entities and constructing a facial dataset of human-like entities; the facial dataset includes images of different degrees of anthropomorphism, and the facial features of the facial image data include at least geometric features, texture features, and hair features;

[0064] In one embodiment, collecting facial image data of a humanoid entity includes:

[0065] By collecting and / or synthesizing facial image data of the humanoid entity; the types of the humanoid entity at least include robots, humanoid models, digital humans, computer-generated characters, real dolls, wax figures and real humans.

[0066] In another embodiment, constructing a facial dataset includes:

[0067] The facial image data of the humanoid entity is preprocessed by using a white background and combining high-resolution frontal portraits and neutral expression images to construct the facial dataset.

[0068] S102. Collect perceptual evaluations of facial data sets, and collect perceptual evaluation data of humanoid entities; the perceptual evaluation data includes anthropomorphism perceptual evaluation data and affinity perceptual evaluation data, and the perceptual evaluation includes geometry, texture, hair anthropomorphism perceptual evaluation, overall anthropomorphism perception, and affinity perceptual evaluation;

[0069] The following is a specific example of steps S101 and S102, which constructs a facial dataset (HumaniFace-1K) to support in-depth research on the uncanny valley effect of humanoid entities. The following are the detailed steps for constructing the dataset:

[0070] (1) Collection of diverse humanoid entity facial image data:

[0071] In order to reflect the widespread existence of humanoid entities in the real world, we synthesized a variety of humanoid images, including robots, humanoid models, digital humans, computer-generated characters, real dolls, wax figures, and real humans. Through these diverse humanoid entities, we ensure that the dataset covers the entire spectrum of anthropomorphism from low to high anthropomorphism. Specifically, the data sources mainly include:

[0072] The dataset contains real-world robot images collected from existing graphics datasets. In addition, more real-world robot images are further collected through the keywords "robot" and "humanoid robot". In order to meet the needs of the dataset, more robot images are also obtained from major modeling websites such as Free3D, Hum3D, and CGTrader.

[0073] To further enrich the dataset, more humanoid images were collected using the keywords "computer-generated characters", "real dolls", "digital humans" and "wax figures". In addition, the facial features of humanoids were divided into geometric features, texture features and detailed features such as hair. These facial attributes were edited and controlled using 3D face reconstruction technology to create humanoids with diverse geometric and texture features, and Blender was used to randomly assign hairstyles to the generated humanoids.

[0074] We also incorporate human face images into our dataset to achieve uniform and dense distribution of anthropomorphism. In addition, we collect more real human face images from databases such as Flickr-Faces High-Quality (FFHQ) and SCUT-FBP5500.

[0075] Image preprocessing:

[0076] In order to ensure the rationality and credibility of the dataset, high-resolution frontal portraits and neutral expression images were combined, and a white background was used to eliminate external interference factors. This processing method ensures high standards in visual quality and consistency for all images, providing strong support for subsequent model training and evaluation. In order to ensure the balance of the dataset, when collecting humanoid entity images, the number of each type of image was controlled to ensure that the distribution of humanoid entities at each level of anthropomorphism in the dataset was relatively balanced, thereby avoiding research bias that may be introduced by sample imbalance. In the end, 1,000 high-quality humanoid entity facial image data were successfully collected.

[0077] (3) Collection of user evaluation data of human-like entities:

[0078] In order to evaluate the anthropomorphism and affinity of humanoid entities, volunteers were recruited for a survey experiment. The 1,000 humanoid entity images in the dataset were divided into 40 groups, each containing 25 images. Participants were randomly assigned to each group and asked to rate the anthropomorphism and affinity of the humanoid entities. Each humanoid entity was rated by at least 30 subjects. Participants used a slider from 1 (extremely dissimilar / extremely unfriendly) to 5 (very similar / very friendly) to indicate their judgment. The subjects' perceived anthropomorphism evaluation of the humanoid entities was collected by asking the subjects "Please judge the following facial geometry / texture / hair and other details / overall anthropomorphism". The subjects' perceived affinity evaluation of the humanoid entities was collected by asking the subjects "Please judge the affinity of the following face".

[0079] (4) Data cleaning:

[0080] During the data cleaning process, a series of measures were taken to ensure the reliability of data analysis. First, duplicates were removed and errors in the data were corrected to ensure the consistency of all data. In addition, data from subjects with incomplete tests, identical humanoid scores, or inappropriate completion times were deleted.

[0081] Thus, through the above steps, various types of humanoid entity images are collected and synthesized, including robots, humanoid models, digital humans, computer-generated characters, real dolls, wax figures, and real humans, ensuring comprehensive coverage of the dataset across the entire anthropomorphic distribution. Each image is processed with high resolution and uses a uniform white background and neutral expression setting to eliminate the interference of external factors on the experimental results. In order to prevent research bias caused by sample imbalance, special attention is paid to the balance of images of different categories during the data collection process to ensure a reasonable distribution of samples at each level of anthropomorphism. The HumaniFace-1K dataset not only solves the problems of limited sample number and single type in existing research datasets, but also provides strong data support for the training of neural network models and the quantitative evaluation of the uncanny valley effect. It is an important basic tool to promote the design and optimization of humanoid entities, and opens up new possibilities for future artificial intelligence applications.

[0082] S103, performing feature quantitative analysis on the perception evaluation data and facial features through a mediation effect model to obtain quantitative results;

[0083] In some embodiments of the present disclosure, the perceived evaluation data and facial features of the humanoid entity are subjected to feature quantification analysis through a mediation effect model to obtain quantified results, including:

[0084] The anthropomorphism perception of geometric features, texture features and hair features in facial features was used as the independent variable, the overall anthropomorphism in the anthropomorphism perception evaluation data was used as the mediating variable, and affinity was used as the dependent variable;

[0085] Using multiple regression analysis to evaluate the first effect of the geometric features, the texture features and the hair features on the affinity, using multiple regression analysis to evaluate the second effect of the geometric features, the texture features and the hair features on the overall anthropomorphism, using univariate regression analysis to evaluate the third effect of the overall anthropomorphism on the affinity, and using the multiplication rule to calculate the indirect effect;

[0086] The Sobels test method was used to evaluate the significance of the indirect effect and determine the mediating effect;

[0087] A comprehensive regression model is constructed by combining the first influence, the second influence and the third influence to evaluate the total influence of the independent variable and the mediating variable on the dependent variable to obtain the quantitative result; wherein the quantitative result includes direct effect and indirect effect.

[0088] In one embodiment, by using the mediation effect model, we systematically explore how the geometric features, texture features, and hair features of humanoid entities directly or indirectly affect human affinity perception. Through this step, we can clarify the specific contribution of various human features to anthropomorphism and affinity, thereby providing a scientific basis for the subsequent optimization of the neural network model. The following are the detailed steps of the mediation effect model:

[0089] The goals of the mediation model are:

[0090] The core goal of the mediation effect model is to explore how different facial features of humanoid entities (such as geometric features, texture features, and hair features) indirectly affect people's affinity perception through overall anthropomorphism, and whether these features have a direct impact on affinity. It is hoped that by quantifying the contribution of each feature to anthropomorphism and affinity, the action paths of these features can be clarified, providing optimization strategies for the design of neural network models.

[0091] Variable definition and selection

[0092] Independent variable(X):

[0093] Geometric features: face shape, facial proportions, contours, etc.

[0094] Texture features: skin smoothness, color, texture details, etc.

[0095] Hair characteristics: hairstyle, hair density, color, etc.

[0096] Mediator variable (M):

[0097] Overall anthropomorphism: Humans' perception of how similar a humanoid entity is to humans overall.

[0098] Dependent variable (Y):

[0099] Affinity: Human's favorability or friendliness towards humanoid entities.

[0100] Assumptions:

[0101] H1: Geometry, texture, and hair features have a significant impact on anthropomorphism (M) (X→M).

[0102] H2: Anthropomorphism has a significant effect on likeability (Y) (M→Y).

[0103] H3: Geometry, texture, and hair features have an indirect effect on likeability through anthropomorphism (X→M→Y).

[0104] H4: Geometry, texture, and hair features have a direct effect on affinity (X→Y).

[0105] Construction of mediation effect model and path analysis:

[0106] Step 1: Test for direct effects (X→Y)

[0107] Multiple regression analysis was used to assess the direct effect of each feature (geometry, texture, hair) on affinity (Y), and the regression model was of the following form:

[0108] Y=β1X1+β2X2+β3X3+∈

[0109] Among them, X1, X2, and X3 represent geometric features, texture features, and hair features, respectively, β1, β2, and β3 represent the regression coefficients of each feature on affinity (i.e., direct effects), and ∈ represents the error term. The regression coefficient β corresponding to each independent variable is estimated using the least squares method. These coefficients represent the direct effect of the independent variable on the dependent variable, independent of the influence of the mediating variable. If β is significantly different from zero (judged by t-test or p-value, p<0.05 is considered significant), it means that the independent variable has a significant direct effect on the dependent variable.

[0110] Step 2: Test the mediation effect path 1 (X→M)

[0111] Multiple regression analysis was used to evaluate the impact of each feature on anthropomorphism (M), and the model form is as follows:

[0112] M=α1X1+α2X2+α3X3+∈

[0113] Among them, α1, α2, and α3 represent the regression coefficients of each feature on anthropomorphism.

[0114] Step 3: Test the mediation effect path 2 (M→Y)

[0115] Univariate regression analysis was used to evaluate the effect of anthropomorphism on likeability, and the model form was as follows:

[0116] Y=γM+∈

[0117] Among them, γ represents the regression coefficient of anthropomorphism on affinity.

[0118] Step 4: Calculate the indirect effect (X→M→Y)

[0119] Calculate the indirect effect using the multiplication rule: Indirect effect = α × γ

[0120] The Sobels test was used to assess the significance of the indirect effect and determine whether there was a significant mediation effect. The Sobels test was used to assess the significance of the indirect effect. Specifically, the Sobels test calculated the standard error of the mediation effect and used a standardized test statistic to calculate the p-value. The steps are as follows:

[0121] Calculate the standard error (SE) of the indirect effect:

[0122]

[0123] Among them, α is the regression coefficient of the independent variable on the mediating variable, γ is the regression coefficient of the mediating variable on the dependent variable,

[0124] and are the standard errors of these regression coefficients, respectively.

[0125] Compute the Sobels test statistic:

[0126]

[0127] The Z value represents the strength of the indirect effect and is equivalent to the standard score in the standard normal distribution. Once the Z value is obtained, the next step is to determine the p-value that corresponds to the Z value. The p-value represents the probability of the observed Z value occurring in the standard normal distribution. The p-value can usually be determined by looking up a standard normal distribution table (Z table) or using statistical software. If the p-value is less than 0.05, it can be considered that the indirect effect is significant, indicating a significant mediation effect.

[0128] Step 5: Comprehensive model analysis (X+M→Y)

[0129] Construct a comprehensive regression model including the independent variables and mediating variables to assess their total effects on the dependent variable:

[0130] Y=β ‘ X+γM+∈

[0131] Among them, β ‘ It is the impact of the independent variable on the dependent variable after taking into account the mediating variables, that is, the adjusted direct effect.

[0132] If in the comprehensive model, β ‘ If γ is still significant and not much different from β in the initial model, it means that the independent variable has a significant direct effect on the dependent variable. If γ is not significant or close to zero, it means that the mediating variable has little effect on the dependent variable and the effect of the independent variable is mainly direct.

[0133] If β ‘ The mediating variable partially explains the influence of the independent variable on the dependent variable, which indicates that there is a significant indirect effect. If the Sobels test verifies that the indirect effect is significant and the direct effect β ‘ If it is not significant or weak, the influence of the independent variable on the dependent variable is mainly indirect.

[0134] Through the above analysis, we can fully utilize the contribution of key features in the neural network model and enhance the model's perception of the uncanny valley effect, thereby optimizing the design of humanoid entities and improving the user's emotional experience.

[0135] Through the mediation effect model, the direct and indirect effects of geometric features (such as face shape and proportion), texture features (such as skin smoothness and color), and hair features (such as hairstyle and hair density) of human-like entities on human affinity perception are systematically explored. The study not only focuses on the contribution of these features to the overall anthropomorphism, but also specifically analyzes whether they indirectly affect affinity through anthropomorphism or have a direct impact on affinity. By conducting a detailed analysis of the impact path of each feature, it is attempted to clarify which features indirectly affect affinity by affecting anthropomorphism and which features directly affect affinity, and quantify their contributions. This dual analysis helps to fully understand the specific mechanism of different features in the uncanny valley effect. After completing these analyses, these findings are applied to the design of neural networks, especially introducing a weighting mechanism in the network architecture so that features that have a greater impact on affinity can obtain higher weights in model training. In addition, by optimizing the model structure, it is ensured that the neural network can fully capture and utilize these key features, thereby significantly improving the network's perception and prediction capabilities of the uncanny valley effect. This solution not only improves the accuracy of the model, but also provides a scientific basis for the design and optimization of humanoid entities, ultimately improving users' emotional experience and acceptance.

[0136] S104, constructing an uncanny valley effect evaluation model for a humanoid entity based on a pre-trained facial feature encoder and the quantization result;

[0137] In one embodiment of the present disclosure, the step of training a facial feature encoder includes:

[0138] Extracting the geometric features, the texture features, and the hair features from a facial image of a humanoid entity;

[0139] During the training process, the output of the encoder is regularized by KL divergence, so that the facial feature encoder generates latent variables with distribution characteristics, wherein the KL divergence is used to measure the difference between the distribution generated by the training encoder and the standard normal distribution.

[0140] In another embodiment of the present disclosure, a model for evaluating the uncanny valley effect of a humanoid entity is constructed based on a pre-trained facial feature encoder and the quantization result, including:

[0141] Based on the quantization result, weighting processing is performed on the extracted geometric features, the texture features and the hair features to generate direct effect weighted features and indirect effect weighted features;

[0142] Dynamically fusing the direct effect weighted features and the indirect effect weighted features through an attention mechanism to obtain a fused feature;

[0143] The fused features are input into two parallel branches for predicting the degree of anthropomorphism and predicting affinity, so as to complete the construction of the uncanny valley effect evaluation model of the humanoid entity.

[0144] In a specific example, Figure 2 As shown in the figure, the direct and indirect effect information obtained from the previous mediation effect analysis is introduced into the network design, so that the neural network can make full use of the weights and relationships of these features to improve the predictive ability of the uncanny valley effect and affinity.

[0145] (1) Humanoid Entity Face Feature Extraction Module

[0146] Assuming the input image is I, the face feature extraction module extracts geometric features, texture features, and hair features from the face image of the humanoid entity. These features describe the shape, surface texture, and appearance characteristics of the face, respectively. The main purpose is to train a high-quality face feature encoder, which is defined as Enc(·), then:

[0147] F geo =Enc geo (I)

[0148] F tex =Enc tex (I)

[0149] F hair =Enc hair (I)

[0150] During the training process, the output of the encoder is regularized by the KL divergence (Kullback-Leibler divergence) to ensure that the encoder can generate hidden variables with good distribution characteristics. The KL divergence is used to measure the difference between the distribution Q(z|x) generated by the training encoder and the standard normal distribution P(z):

[0151]

[0152] Through these trainings, feature extractors that can learn high-quality humanoid entity geometry, texture, and hair are trained. These feature extractors can provide the basis for subsequent feature fusion and prediction.

[0153] (2) Mediating effect weighting module:

[0154] Based on the results of the mediation effect analysis, the extracted geometric, texture and hair features are weighted to generate features with weighted direct and indirect effects. The direct effect weighting is used to capture the direct impact of the feature on affinity, while the indirect effect weighting is used to describe the indirect impact of the feature on affinity through anthropomorphism.

[0155] Direct effect weighting:

[0156] F direct =β geo,direct ·F geo +β tex,direct ·F tex +β hair,direct ·F hair

[0157] Among them, β geo,direct , β tex,direct and β hair,direct Respectively represent the direct effect weights of geometry, texture, and hair features in the mediation effect results.

[0158] Weighted indirect effect (through the influence of anthropomorphism):

[0159] F indirect =(α geo ×γ)·F geo +(α tex ×γ)·F tex +(α hair ×γ)·F hair

[0160] Among them, α geo , α tex and α hair is the influence weight of each feature on anthropomorphism, and is the influence weight of anthropomorphism on affinity.

[0161] (3) Feature fusion module of attention mechanism:

[0162] The weighted features are dynamically fused through the attention mechanism to optimize the comprehensive representation of the features. The attention mechanism enables the model to automatically adjust the importance of each feature to ensure the maximum utilization of effective information during the feature fusion process. The attention mechanism plays a key role in feature fusion, enabling the model to dynamically adjust the importance of each feature according to the context. Through the cross-attention and self-attention mechanisms, the model can more intelligently fuse the direct and indirect weighted features to generate a more effective comprehensive feature representation. The formula for the cross attention mechanism is:

[0163]

[0164] The formula of the self-attention mechanism is expressed as:

[0165]

[0166] Among them, Q, K, V are the weighted feature maps of the input, and dk is the feature dimension.

[0167] Feature Fusion:

[0168]

[0169] Anthropomorphism and affinity prediction module:

[0170] The fused feature map F fused is fed into two parallel branches, one for predicting anthropomorphism and the other for predicting affinity. The anthropomorphism prediction is:

[0171]

[0172] in, is the predicted anthropomorphism value, Conv M is the convolution operation, FC M It is a fully connected layer.

[0173] Affinity prediction is:

[0174]

[0175] in, is the predicted affinity value, Conv Y is the convolution operation, FC Y It is a fully connected layer.

[0176] S105, evaluating the facial image of the humanoid entity using the uncanny valley effect evaluation model of the humanoid entity;

[0177] In another embodiment of the present disclosure, it further includes:

[0178] Introducing a multi-task loss function to optimize the uncanny valley effect evaluation model of the humanoid entity;

[0179] The definition of the multi-task loss function includes:

[0180] In a specific example, in order to optimize the model's anthropomorphism and affinity prediction, a multi-task loss function was designed to comprehensively consider the prediction errors of anthropomorphism and affinity.

[0181] Anthropomorphism loss:

[0182]

[0183] Where N is the number of samples, M is i is the true anthropomorphism value.

[0184] Affinity loss:

[0185]

[0186] Among them, Yi is the true anthropomorphism value.

[0187] The total loss function is:

[0188] L=λ1L M +λ2L Y

[0189] Among them, λ1 and λ2 are weight hyperparameters that control the two loss terms. The best value is obtained by tuning the validation set. The Adam optimizer is selected to minimize the loss function. The initial learning rate α is usually set to 10 -4 , and dynamically adjust according to the training effect. In order to prevent the model from overfitting, Dropout regularization is introduced in the fully connected layer, and L2 regularization terms are added to the loss function. Standard regression evaluation indicators (such as mean square error MSE, standard mean square error RMSE, determination coefficient R 2 Evaluate the performance of the model on the validation set (etc.), focusing on the model’s prediction accuracy of anthropomorphism and affinity, and its ability to perceive the uncanny valley effect.

[0190] An optimized neural network model was constructed based on the analysis results of the mediation effect model. First, specialized encoders were trained to extract geometric, texture, and hair features from facial images of humanoid entities. These encoders were regularized with KL divergence to ensure that the extracted features had stable distribution characteristics. Next, these extracted features were input into the feature weighting module, and the weights of each feature were dynamically adjusted according to the contribution of geometric, texture, and hair features to anthropomorphism and affinity obtained from the mediation effect analysis. After that, the attention mechanism was used for feature fusion, and the interaction between features was enhanced through cross-attention and self-attention mechanisms, so that the model could better capture the information of key features. Finally, the fused features were input into the parallel prediction module, and the scores of anthropomorphism and affinity were output respectively. By designing a multi-task loss function, the model was effectively optimized in the prediction of anthropomorphism and affinity, thereby improving the model's perception of the uncanny valley effect and the accuracy of affinity evaluation. This neural network structure constructed based on the mediation effect results provides solid technical support for the design and optimization of humanoid entities.

[0191] In summary, this embodiment proposes a method for evaluating the uncanny valley effect of humanoid entities. Dataset construction (HumaniFace-1K): First, a comprehensive facial dataset of HumaniFace-1K is constructed, which covers a wide range of humanoid entity images from low to high anthropomorphism, ensuring the diversity and scientificity of the data. Through high-resolution image processing, standardized background settings, and fine facial feature annotation, the quality and representativeness of the dataset are ensured, providing a solid foundation for subsequent analysis and model training.

[0192] Based on the data set, the mediation effect model was used to systematically analyze the impact paths of geometric features, texture features, hair features, etc. on human affinity perception. The mediation effect model helped clarify whether these features indirectly affect affinity through overall anthropomorphism or have a direct impact on affinity, and quantified the contribution of each feature. These analysis results provide key references for subsequent neural network optimization.

[0193] Based on the analysis results of the mediation effect model, a neural network model HumaniNet was designed and optimized. Specifically, a facial feature encoder was trained to extract geometric, texture, and hair features from the facial images of humanoid entities and ensure that these features have a stable distribution. Then, based on the results of the mediation effect analysis, these features were weighted to adjust their importance in the model. Next, the weighted features were fused using the attention mechanism to enhance the interaction between features and improve the model's ability to capture key information. Finally, the fused features were input into the prediction module to generate scores for anthropomorphism and affinity, respectively. Through the optimization of the multi-task loss function, the model can perceive the uncanny valley effect more accurately and provide technical support for the design and optimization of humanoid entities.

[0194] The uncanny valley effect assessment method for humanoid entities provided in this embodiment can be executed in a smart terminal, a computer terminal, a network device, a chip, a chip module or a similar computing device.

[0195] Corresponding to the above-described method for evaluating the uncanny valley effect of a humanoid entity, this embodiment further provides a system for evaluating the uncanny valley effect of a humanoid entity. The following will introduce each of them. Specifically, Figure 3 As shown, the uncanny valley effect assessment system of a humanoid entity in this embodiment includes:

[0196] The data set establishment module 1 is used to collect facial image data of humanoid entities and construct a facial data set of humanoid entities; the facial data set of humanoid entities includes images with different degrees of anthropomorphism, and the facial features of the facial image data include at least geometric features, texture features and hair features;

[0197] Evaluation module 2 is used to collect perceptual evaluations of facial data sets and collect perceptual evaluation data of humanoid entities; the perceptual evaluation data includes anthropomorphism perceptual evaluation data and affinity perceptual evaluation data, and the perceptual evaluation includes geometry, texture, hair anthropomorphism perceptual evaluation, overall anthropomorphism perception and affinity perceptual evaluation;

[0198] The mediation effect model module 3 is used to perform feature quantitative analysis on the perception evaluation data and facial features through the mediation effect model to obtain quantitative results;

[0199] Model building module 4, used to build an uncanny valley effect evaluation model of humanoid entities based on the pre-trained facial feature encoder and quantization results;

[0200] Evaluation module 5 evaluates the facial image of the humanoid entity using the uncanny valley effect evaluation model of the humanoid entity.

[0201] It should be noted that the uncanny valley effect evaluation system embodiment of a humanoid entity provided in the embodiments of the present application and the uncanny valley effect evaluation method embodiment of a humanoid entity described above are based on the same inventive concept.

[0202] It should be noted that the uncanny valley effect assessment system for humanoid entities in this embodiment can be, for example: a separate chip, a chip module or an electronic device, or a chip or chip module integrated into an electronic device. Regarding the various modules / units contained in the various devices and products described in the above embodiments, they can be software modules / units, or hardware modules / units, or they can be partially software modules / units and partially hardware modules / units. For example, for various devices and products applied to or integrated in a chip, the various modules / units contained therein can all be implemented in the form of hardware such as circuits, or at least some of the modules / units can be implemented in the form of software programs, which run on a processor integrated inside the chip, and the remaining (if any) modules / units can be implemented in the form of hardware such as circuits; for various devices and products applied to or integrated in a chip module, the various modules / units contained therein can all be implemented in the form of hardware such as circuits, and different modules / units can be located in the same component (such as a chip, circuit module, etc.) or different components of the chip module, or at least some of the modules / units can be implemented in the form of software programs, which run on a processor integrated inside the chip, and the remaining (if any) modules / units can be implemented in the form of hardware such as circuits; for various devices and products applied to or integrated in a chip module, the various modules / units contained therein can all be implemented in the form of hardware such as circuits, and different modules / units can be located in the same component (such as a chip, circuit module, etc.) or different components of the chip module, or at least some of the modules / units can be implemented in the form of hardware such as circuits. The element can be implemented in the form of a software program, which runs on a processor integrated inside the chip module, and the remaining (if any) modules / units can be implemented in the form of hardware such as circuits; for various devices and products applied to or integrated in the terminal, the various modules / units contained therein can be implemented in the form of hardware such as circuits, and different modules / units can be located in the same component (for example, chip, circuit module, etc.) or in different components in the terminal, or, at least some modules / units can be implemented in the form of a software program, which runs on a processor integrated inside the terminal, and the remaining (if any) modules / units can be implemented in the form of hardware such as circuits.

[0203] A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements a method for evaluating the uncanny valley effect of a humanoid entity when executing the program. Figure 4 The electronic device 30 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0204] like Figure 4 As shown, the electronic device 30 may be in the form of a general-purpose computing device, for example, it may be a server device. The components of the electronic device 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including the memory 32 and the processor 31).

[0205] The bus 33 includes a data bus, an address bus, and a control bus.

[0206] The memory 32 may include a volatile memory, such as a random access memory (RAM) 321 and / or a cache memory 322 , and may further include a read-only memory (ROM) 323 .

[0207] The memory 32 may also include a program / utility 325 having a set (at least one) of program modules 324, such program modules 324 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0208] The processor 31 executes various functional applications and data processing by running the computer program stored in the memory 32, such as the uncanny valley effect assessment method of a humanoid entity in Example 1 of the present invention.

[0209] The electronic device 30 may also communicate with one or more external devices 34 (e.g., keyboards, pointing devices, etc.). Such communication may be performed via an input / output (I / O) interface 35. Furthermore, the model generating device 30 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 36. Figure 4 As shown, the network adapter 36 communicates with other modules of the model-generated device 30 via the bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the model-generated device 30, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.

[0210] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0211] This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for evaluating the uncanny valley effect of a humanoid entity.

[0212] The readable storage medium may include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device or any suitable combination of the above.

[0213] In a possible implementation, the present invention can also be implemented in the form of a program product, which includes program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps in the method for evaluating the uncanny valley effect of a humanoid entity.

[0214] Among them, the program code for executing the present invention can be written in any combination of one or more programming languages, and the program code can be executed completely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or completely on the remote device.

[0215] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this embodiment is not limited here.

[0216] It should be understood that the "system", "device", "unit" and / or "module" used in this embodiment is a method for distinguishing different components, elements, parts, parts or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0217] In this embodiment, a flow chart is used to illustrate the operations performed by the system according to the embodiment of the present embodiment. It should be understood that the preceding or following operations are not necessarily performed precisely in order. On the contrary, each step can be processed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0218] As shown in this embodiment, unless the context clearly indicates an exception, the words "a", "an", "a kind" and / or "the" do not specifically refer to the singular, but may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate that the steps and elements that have been clearly identified are included, and these steps and elements do not constitute an exclusive list, and the method or device may also include other steps or elements.

[0219] The definition of including in this embodiment, as used herein, the terms "having", "may have", "including", or "may include" indicate the existence of the corresponding functions, operations, elements, etc. of this embodiment, and do not limit the existence of one or more other functions, operations, elements, etc. In addition, it should be understood that the terms "including" or "having" as used herein indicate the existence of the features, numbers, steps, operations, elements, parts, or a combination thereof described in the specification, and do not exclude the existence or addition of one or more other features, numbers, steps, operations, elements, parts, or a combination thereof.

[0220] In the definition of and / or in this embodiment, the term "A or B", "at least one of A and / or B" or "one or more of A and / or B" as used herein includes any and all combinations of the words listed therewith. For example, "A or B", "at least one of A and / or B" or "one or more of A and / or B" means (1) including at least one A, (2) including at least one B, or (3) including both at least one A and at least one B.

[0221] The definitions of first and second in this embodiment, and the descriptions of first, second, etc. in this embodiment are only used for illustration and distinction of the described objects, and have no order, nor do they represent a special limitation on the number of devices in this embodiment, and cannot constitute any limitation on this embodiment. For example, a first element may be referred to as a second element without departing from the scope of this disclosure, and similarly, a second element may be referred to as a first element.

[0222] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that this is only for illustration and the protection scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A method for evaluating the uncanny valley effect of a humanoid entity, characterized in that: include: Collecting facial image data of humanoid entities and constructing a facial dataset of humanoid entities; the facial dataset of humanoid entities includes images of different degrees of anthropomorphism, and the facial features of the facial image data include at least geometric features, texture features and hair features; Soliciting perceptual evaluations of the facial data set, and collecting perceptual evaluation data of the humanoid entity; the perceptual evaluation data includes anthropomorphism perceptual evaluation data and affinity perceptual evaluation data, and the perceptual evaluation includes geometry, texture, hair anthropomorphism perceptual evaluation, overall anthropomorphism perception, and affinity perceptual evaluation; Performing feature quantification analysis on the perception evaluation data and the facial features through a mediation effect model to obtain a quantified result; Constructing an uncanny valley effect evaluation model for humanoid entities based on a pre-trained facial feature encoder and the quantization results; The uncanny valley effect evaluation model of the humanoid entity is used to evaluate the facial image of the humanoid entity.

2. The method for evaluating the uncanny valley effect of a humanoid entity as claimed in claim 1, characterized in that: The collecting of facial image data of a human-like entity comprises: By collecting and / or synthesizing facial image data of the humanoid entity; the types of the humanoid entity at least include robots, humanoid models, digital humans, computer-generated characters, real dolls, wax figures and real humans.

3. The method for evaluating the uncanny valley effect of a humanoid entity as claimed in claim 2, characterized in that: The constructing of the facial dataset comprises: The facial image data of the humanoid entity is preprocessed by using a white background and combining high-resolution frontal portraits and neutral expression images to construct the facial dataset.

4. The method for evaluating the uncanny valley effect of a humanoid entity as claimed in claim 1, characterized in that: The step of performing feature quantification analysis on the perception evaluation data of the humanoid entity and the facial features through the mediation effect model to obtain quantification results includes: The anthropomorphism perception of the geometric features, the texture features and the hair features in the facial features is used as an independent variable, the overall anthropomorphism in the anthropomorphism perception evaluation data is used as a mediating variable, and affinity is used as a dependent variable; Using multiple regression analysis to evaluate the first effect of the geometric features, the texture features and the hair features on the affinity, using multiple regression analysis to evaluate the second effect of the geometric features, the texture features and the hair features on the overall anthropomorphism, using univariate regression analysis to evaluate the third effect of the overall anthropomorphism on the affinity, and using the multiplication rule to calculate the indirect effect; The Sobels test method was used to evaluate the significance of the indirect effect and determine the mediating effect; A comprehensive regression model is constructed by combining the first influence, the second influence and the third influence to evaluate the total influence of the independent variable and the mediating variable on the dependent variable to obtain the quantitative result; wherein the quantitative result includes direct effect and indirect effect.

5. The method for evaluating the uncanny valley effect of a humanoid entity as claimed in claim 1, characterized in that: The steps to train the facial feature encoder include: Extracting the geometric features, the texture features, and the hair features from a facial image of a humanoid entity; During the training process, the output of the encoder is regularized by KL divergence, so that the facial feature encoder generates latent variables with distribution characteristics, wherein the KL divergence is used to measure the difference between the distribution generated by the training encoder and the standard normal distribution.

6. The method for evaluating the uncanny valley effect of a humanoid entity as claimed in claim 1, characterized in that: The uncanny valley effect evaluation model of a humanoid entity is constructed based on the pre-trained facial feature encoder and the quantization result, comprising: Based on the quantization result, weighting processing is performed on the extracted geometric features, the texture features and the hair features to generate direct effect weighted features and indirect effect weighted features; Dynamically fusing the direct effect weighted features and the indirect effect weighted features through an attention mechanism to obtain a fused feature; The fused features are input into two parallel branches for predicting the degree of anthropomorphism and predicting affinity, so as to complete the construction of the uncanny valley effect evaluation model of the humanoid entity.

7. The method for evaluating the uncanny valley effect of a humanoid entity as claimed in claim 1, characterized in that: Also includes: Introducing a multi-task loss function to optimize the uncanny valley effect evaluation model of the humanoid entity; The definition of the multi-task loss function includes: Anthropomorphism loss: Where N is the number of samples, M is i is the true anthropomorphism value; Affinity loss: Among them, Y is the true anthropomorphism value; The total loss function is: L=λ1L M +λ2L Y Among them, λ1 and λ2 are weight hyperparameters that control the two loss terms.

8. A system for evaluating the uncanny valley effect of a humanoid entity, characterized in that: include: A data set establishment module, used to collect facial image data of humanoid entities and construct a facial data set of humanoid entities; the facial data set of humanoid entities includes images of different degrees of anthropomorphism, and the facial features of the facial image data include at least geometric features, texture features and hair features; An evaluation module, used to collect perceptual evaluations of the facial data set, and collect perceptual evaluation data of the humanoid entity; the perceptual evaluation data includes anthropomorphism perceptual evaluation data and affinity perceptual evaluation data, and the perceptual evaluation includes geometry, texture, hair anthropomorphism perceptual evaluation, overall anthropomorphism perception and affinity perceptual evaluation; A mediating effect model module, used to perform feature quantification analysis on the perception evaluation data and the facial features through a mediating effect model to obtain a quantified result; A model building module, used to build an uncanny valley effect evaluation model of a humanoid entity based on a pre-trained facial feature encoder and the quantization result; An evaluation module evaluates the facial image of the humanoid entity using the uncanny valley effect evaluation model of the humanoid entity.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images